• VirtualDJ now supports OneLibrary exports to compatible DJ hardware. Full compatibility between applications still isn’t guaranteed.

• Serato’s 5.0 beta lets streaming tracks and local files share a crate. That doesn’t give them the same playback or export options.

• Before changing software or taking a USB to unfamiliar gear, test a small playlist. Check the audio, cue points and beatgrids on the destination setup.

A track loads on another player, but the cue point you use to start your mix is missing. Or the beatgrid you corrected is wrong again. The music made it across. Now you have preparation to redo.

That’s the problem I care about with DJ library compatibility. Getting the audio onto another system is only part of the job. I also want to know whether the work I’ve already done will be there when I load the track.

Two recent announcements address different parts of this. Serato is making it easier to organize local files and streamed music together. OneLibrary is making it easier to export prepared collections to compatible gear. Neither removes the need to check what your particular setup supports.

A shared crate doesn’t mean shared access

Serato announced its DJ 5.0 beta on September 10. Its Open Streaming feature lets you sign into multiple supported streaming services at once and put tracks from those services in the same crate as your local files. A crate is Serato’s way of organizing a collection of tracks.

That’s convenient when you’re building a set from several sources. But a local file and a streaming entry still depend on different things. With a local file, you have the audio stored on your drive. With a streamed track, playback depends on the service and the options its DJ integration supports.

rekordbox’s Spotify integration shows why this distinction matters. Version 7.2.18 made Spotify tracks available in EXPORT mode, the part of rekordbox used to prepare music for compatible players. Yet the release notes explicitly say those tracks can’t be loaded or played on CDJ/XDJ players or all-in-one systems through USB EXPORT or LINK EXPORT.

USB EXPORT prepares a drive you take to the player. LINK EXPORT lets a compatible player load music from rekordbox over a connection to your computer. Seeing a Spotify track in EXPORT mode doesn’t mean either route can play it. Supported hardware may offer another way to access Spotify, but that’s a separate playback method.

Check that method before you build a set around it.

Streaming changes can affect preparation too. In February, rekordbox warned that a change in SoundCloud’s audio format could misalign existing cue points and beatgrids. Affected tracks could need reanalysis or their cue points reset. You don’t have to switch applications for preparation to need checking again.

Serato also recommends keeping your existing version installed and avoiding its beta for critical performances. Try the new features somewhere a failure won’t interrupt a gig.

What OneLibrary can carry

On September 15, AlphaTheta announced VirtualDJ support for OneLibrary. A supported version of VirtualDJ can export its collection to a USB drive in that format for use on compatible DJ hardware.

AlphaTheta says OneLibrary exports can include music files, playlists, cue points, beatgrid information and waveforms. Cue points mark places you want to return to. Beatgrids map the beats so the software or player can follow the track’s timing.

But AlphaTheta also says full compatibility between DJ applications isn’t guaranteed. Export specifications can differ by application. Don’t assume every saved loop, comment or correction will transfer just because both products support OneLibrary.

Exporting a playlist to a player and importing it into another DJ application are different tests. A successful USB export tells you what worked on that player. It doesn’t establish what another application will import.

The player matters too. AlphaTheta documents two USB library formats: OneLibrary and the older Device Library. Its March notice describes cases where playlists appear on one generation of player but not another, with conversion sometimes needed. Firmware can also affect which library format a player reads.

Before switching software or playing on unfamiliar equipment, check the export format, the exact player model and firmware and the preparation data you rely on.

Test one playlist before moving everything

Choose a small playlist you’d actually play. Twenty tracks is a reasonable starting point. Include a track with a manually corrected beatgrid and another with cue points or saved loops you use.

Decide what you’re testing: a USB export to a player or a move into another DJ application. Write down the software version, export format and destination. For hardware, include the model and firmware version. If the test works, you’ll know which setup you tested.

Before exporting, note where the audio comes from. Separate files stored on your drive from tracks supplied by a streaming service. For streamed tracks, check the documented playback requirements for the software and hardware you intend to use.

For a USB test, export the playlist through your normal process and load it on the destination player. For an application move, use the documented import or conversion process. Keep a backup of your original library before trying either.

Check:

• Can you find and load every track?

• Are your cue points at the right positions?

• Does the manually corrected beatgrid still line up?

• Is the playlist order intact?

• Did the saved loops and comments you rely on transfer?

Play through the sections where those details matter. A cue marker can look present without being where you expect it. A grid deserves more than a glance at the first beat.

If something’s missing, work out whether the destination doesn’t support it or whether you need to change the export process. Do that with a small playlist before committing your whole collection.

If you’re also changing how the players connect, I covered the network side in Your DJ Booth Is a Network Now⁠. Library compatibility and the connection carrying the music both need to work.

Back up the preparation too

A backup of your music folder protects the files in that folder. It doesn’t necessarily include the library database or other files where your DJ application stores preparation.

The backup procedure depends on the application. Serato’s instructions identify the music files, the _Serato_ folders and a separate Library folder. It can create _Serato_ folders on external drives too, so check those rather than assuming everything lives on your computer.

rekordbox provides a library backup and restore process that includes your collection, playlists, cues and analysis information. It also gives you the option to back up the music files.

Use the official instructions for your application and version. A USB prepared for performance and a backup you can restore into your application serve different purposes. Know which one you’ve made.

Test the restore process on a separate setup or with the existing library preserved. Check a familiar playlist, a corrected grid and the cues you actually use. For streamed music, also check what remains accessible if the service or account is unavailable.

OneLibrary may make a move easier. I’d still want to load a familiar track on the destination setup and check the details myself.

Which part of your prepared library have you actually tested on another system?

Resources

• AlphaTheta: VirtualDJ officially supports OneLibrary⁠, September 15, 2026.

• AlphaTheta: OneLibrary features and compatibility FAQ⁠.

• Serato: Lucid Beatgrids Just Dropped in the Serato DJ 5.0 Beta⁠, September 10, 2026.

• rekordbox: Version 7.2.18 release information⁠, August 18, 2026.

• rekordbox: Version 7.2.10 release information⁠, February 3, 2026. Includes the warning about SoundCloud cue points and beatgrids.

• AlphaTheta: Important Notice for Customers Using USB Devices with Our DJ Equipment⁠, March 18, 2026.

• Serato: Backing up and moving your library to a new computer⁠.

• rekordbox: Library backup and transfer FAQ⁠.

• I rebuilt Forest Blakk’s “Long Story Short” using only his vocal from the original recording.

• In the first version, I moved vocal phrases to fit my arrangement. Some edits changed how the performance felt to me.

• Next, I’ll keep the original phrase order and spacing, then build the music around them. This is an unofficial reimagining, not an official Forest Blakk remix or release. Forest’s team reviewed the article and gave me permission to make the track public on SoundCloud.

  

I’d planned an eight-bar build, where the music gathers tension before the drop. Forest’s vocal phrase ran past it. In Ableton Live 11, I could’ve trimmed the phrase or moved part of it to make the drop arrive where I’d planned.

Instead, I extended the build to ten bars so Forest could finish the phrase as his delivery grew more intense.

This is an unofficial reimagining of Forest Blakk’s “Long Story Short,” not an official Forest Blakk remix or release. Forest’s team initially declined release clearance, but after reviewing this article they gave me permission to make the track public on SoundCloud. René Renner, COO of Logan Media Entertainment, specifically asked that I make clear this is not an official remix.

I started with one rule: keep the voice

My wife introduced me to “Long Story Short,” and we both connected with it. I wanted to keep Forest’s vocal and write a new melodic, progressive electronic arrangement around it.

His vocal was the only audio I carried over from the original recording. I built the drums, bass and chord parts myself.

But I’d left myself one freedom: I could move the vocal phrases.

I cut phrases, rearranged sections and reused vocal fragments. I still wanted the lyrics to make emotional sense, but I could change when a line arrived or how long the gap before it lasted to fit my arrangement.

That gave me room to build the track. But it left me wondering how much I was changing the feeling of the performance by changing its pauses and timing.

The phrase that broke the eight-bar habit

The ten-bar build was one place where I’d already let the vocal change my plan. Cutting the phrase at the eight-bar mark felt abrupt, so I delayed the drop and let the phrase finish.

As the snare roll built toward the drop, it became harder to hear the vocal clearly. The phrase had enough time now, but it needed more room in the mix.

I automated the vocal volume to rise through the build so it stayed audible as the snare roll intensified. At the drop, I brought the vocal level back down.

Here, the volume automation kept attention on the vocal as the build got busier.

I also added delay throws: echoes on selected phrase endings. They filled the gaps without leaving echoes running over the next vocal line.

The vocal was now shaping both the section length and how I balanced the build around it.

The vocal contains more than lyrics

I used the vocal melody to guide the new chords and instrumental parts. I built my arrangement in A major, with F-sharp minor, D, A and E chords through the verses. In the drops, I used A, E, F-sharp minor and D, with a D-to-E chord change leading into the main drop.

In the second drop, the lead instrument plays Forest’s chorus melody two octaves above the vocal. Another lead later adds a harmony a third above it, using notes from A major. I wrote new instrumental parts around the melody already present in his singing.

I also changed the pitch range of an arpeggiated part, a repeating pattern that plays chord notes one at a time. I moved it higher for a brighter sound and lower where that helped the vocal stay clear. In the bigger drop, I changed its range again to make room for the lead melody.

Sometimes changing the notes’ pitch range separates two parts more clearly than adjusting their tone with EQ.

Next, I’ll keep the vocal timing intact

For the next version, I’ll use the same vocal source and build all the instrumental parts again. This time, I’ll keep the phrases in their original order and preserve the pauses between them. I’ll decide where sections begin and end around that performance.

That might mean a seven-bar section instead of eight, or a transition later than I’d usually place it. I may leave a pause open where I’d normally add a fill. The section lengths will follow the vocal rather than my usual pattern.

I haven’t made the second version yet, so I don’t know how it’ll compare.

I expect keeping the original pauses and phrase lengths to preserve more of the feeling I heard in Forest’s performance. I also expect it to be harder to build a steady electronic groove around it. Repeated patterns and familiar section lengths help establish when a change feels due. I want to hear what happens when the vocal sets that timing.

I want to compare how the vocal feels and how the track moves in each version. Keeping the original timing won’t automatically make the better track.

Try both approaches with a vocal you have permission to edit

Start with a vocal you recorded yourself or have permission to edit. Make two copies of the same starting session.

In version A, set the section lengths and major transitions first. Then move vocal phrases or shorten the gaps between them to fit.

In version B, keep the vocal phrases in their original order with the original gaps between them. Build the drums, chords and transitions around that timing.

Then listen to both without looking at the session timeline. Does a transition cut off a phrase before it feels finished? Do the pauses help, or does the track lose momentum? Listen for drum fills that make a line harder to follow, and whether the unusual section lengths sound unusual when you can’t see the bar numbers.

You may prefer the version with regular section lengths. I might too. That’s why I want to try both.

In the first version, I could rearrange a lot of the vocal and still feel connected to the performance. But the phrase in that build felt cut short until I gave it two more bars.

Next, I’ll leave the vocal timing intact and see what I have to change around it.

In your own session, what would you have to change if you couldn’t move a vocal phrase or shorten a pause?

Resources

• Forest Blakk, “Long Story Short” on Spotify, official recording reference.

• ZenOne, “Long Story Short” unofficial reimagining on SoundCloud - public SoundCloud track shared with permission from Forest Blakk’s team. Not an official Forest Blakk remix or release.

• Suno says v6 can now change one section or one lyric while preserving the rest of a song. Studio 2.0 adds MIDI, effects, automation and a chat layer inside a browser-based DAW.

• I don’t use Suno to make ZenOne music. My releases are human-made. I’m looking at this because targeted AI editing creates a useful producer test whether you use these tools or not.

• That test is edit locality: change one thing, then listen closely to everything that was supposed to stay the same (regression).

In June, I wrote something that already needs a footnote: generation is addition by nature.

The point in The Sound You Can’t Remove was that a generator has no fear of filling an empty bar. A producer’s harder job is often subtraction: deciding what the track can lose without losing itself.

That distinction is getting less clean.

I don’t use Suno in my ZenOne production. My releases are human-made. I’m looking at Suno here because its new editing model raises a producer question I think matters even if you never put a generative tool in your workflow: when you change one thing, what else moves?

On September 9, Suno released v6. Its own release notes say the model can change one section of an existing song while preserving the rest. It can replace a single lyric without rebuilding the whole track, combine parts from multiple songs, isolate a sample and build around it. A month earlier, Suno Studio 2.0 added MIDI editing, effects, automation, a wavetable synth and a chat layer to its browser-based production environment.

I don’t think the interesting part is that AI can make another song faster. We already crossed that line.

The interesting part is that the generator is learning to take a note.

The first draft was never the hard part

A first pass can be impressive and still be useless to a producer.

You hear eight bars and like the atmosphere, but the kick is wrong. The vocal works until the last word. The breakdown is good, but the transition into it gives away the tension too early. In a normal DAW, those aren’t reasons to start over. They’re reasons to edit.

Production is full of narrow decisions. Move this note. Shorten that reverb tail. Pull one percussion hit forward. Keep the bass exactly where it is.

That last sentence matters.

Suno’s v6 announcement is interesting because it makes preservation part of the product claim. The example isn’t simply “make a different chorus.” It’s change the chorus while keeping the rest of the song. Studio’s chat documentation pushes in the same direction: select part of a project, describe the change and let the system perform operations inside the multitrack session.

That’s much closer to production than one-shot generation is.

It also creates a much harder standard.

The new test is edit locality

I want a term for the thing producers should listen for here, so I’ll use edit locality.

Edit locality means the change stays where you asked for it.

If you ask for a different snare and the snare changes, that’s obvious. But did the groove around it move? Did the room sound change? Did the bass transient soften? Did the stereo image shift? Did a tiny vocal timing detail you liked get regenerated into something cleaner but less interesting?

Those changes might be subtle enough that the new version still sounds “good.” That’s not the same as being a precise edit.

A useful editor has to preserve context.

This is why the move toward generative DAWs matters more to me than another leap in first-pass fidelity. Studio 2.0 now has familiar production objects: MIDI clips, effects chains, automation lanes, synth parameters and a timeline. Suno says its chat layer can see what you have selected and make changes using operations available inside the project.

Those are all ways of narrowing the blast radius of a decision.

But the interface can look precise before the result actually is. A highlighted region on a timeline doesn’t prove everything outside that region remained musically identical. A text instruction that says “leave the bass alone” isn’t the same thing as a locked bass track.

The ears still have to verify the boundary.

Run the one-change test

If you use Suno or another generative music tool with targeted editing, don’t evaluate it by asking for the best song it can make. That tells you almost nothing about whether it can function as an editor.

Give it a smaller problem.

Start with a piece where you already know what you want to keep. It can be something generated inside the system or material you’re allowed to use there. Pick one exact revision. Not “make it better.” Something like:

1. Change the snare in bars 17 through 24.

2. Replace one lyric and keep the vocal phrasing around it.

3. Make the pad darker without changing the bass or drums.

4. Shorten one transition without changing what follows it.

Export the before and after versions. Match their playback level closely enough that louder doesn’t win by accident. Then ignore the part you asked to change for a moment.

Listen to everything else.

Did the kick keep the same weight? Does the bass enter on the same micro-timing? Did the vocal consonants around the edit move? Is the ambience the same? Did a cymbal tail disappear? Does the groove lean differently even though you never asked it to?

Then do the same edit again.

If the preserved material moves differently on every attempt, you’ve learned something important. The system may be a good generator with an editing interface around it. If the requested change stays local and the rest remains stable, that’s a different class of tool.

The score isn’t “version B sounds better.”

The score is how much of version A you were able to keep on purpose.

Control is what stays still

The shift is wider than Suno. Universal Music Group and ElevenLabs announced a licensed AI music platform on September 10 that’s intended to let fans make remixes, mashups and new interpretations from participating artists’ music. Suno’s v6 announcement talks about mashups, sampling and targeted revision. Studio is becoming more like a DAW.

Different products, different rights models, but the direction rhymes. AI music is moving from “give me a song” toward “work on this material.”

That changes the argument producers should be having.

The old debate was mostly about generation quality and authorship. Those questions aren’t going away. None of this changes my own workflow: I don’t use Suno to make ZenOne music. My releases are written, arranged, sound-designed, produced and mixed by me. They’re human-made, and if I ever use generative AI in the audio, composition, lyrics or vocals, I’ll say so explicitly.

But there’s another question now: what does control mean when the tool can revise instead of only generate?

I don’t think control means having more prompts. It means being able to make a narrow musical decision and trust the boundaries around it.

That’s familiar territory for anyone who works in a DAW. We freeze tracks because we don’t want them changing. We duplicate before destructive edits. We automate one parameter while leaving the others alone. We save versions because revision is partly the art of deciding what must remain true.

The generator becoming an editor doesn’t make that discipline less important.

It makes it measurable.

Ask the tool to change one thing. Then listen hardest to everything you told it not to touch.

Leave a comment

Resources

• Suno, “Introducing v6,” September 9, 2026: targeted section edits, single-lyric changes, mashups, sampling and multimodal creation.

• Suno, “Studio 2.0 is here,” August 13, 2026: MIDI, audio effects, wavetable synth, musical typing, automation and chat inside Suno Studio.

• Suno, “Updates to Suno Studio,” September 2, 2026: BPM-aware chat, undo for prompt edits, plugin workflow improvements and performance fixes.

• Suno, “Introducing Studio 2.0,” August 13, 2026: additional detail on MIDI as a prompt source and the Studio chat workflow.

• Universal Music Group, UMG and ElevenLabs strategic agreement, September 10, 2026: announced licensed platform for remixes, mashups and new interpretations using participating artists’ music.

• A vocal recording doesn’t have to remain recognizable as singing. Its breath, changing pitch or sharp beginning can become material for another instrument.

• Producer Ital Tek describes this approach in a recent interview. In my track Anamorph, one vocal recording became seven different musical parts.

• Try it with your own voice. Judge the result inside a track before deciding which rough edges to fix.

The recording is a little awkward. The note wobbles, the breath is louder than expected and the ending trails off before you’ve decided what to do with it. If you want to use it as the sung part of a song, you might record another take. If you plan to cut it up or stretch it into something else, you haven’t heard what it can do yet.

Subscribe now

Ital Tek turned an acoustic-guitar recording into a kick-drum sound. His label, Planet Mu, describes it on “Kill Switch,” a track from Mind Abandon, released in May. The guitar supplied the recorded sound. Processing turned it into percussion.

In his September 2 interview with Ableton, Alan Myson, who records as Ital Tek, describes making quick recordings of his voice and instruments before knowing where they’d fit in a song. He then worked with those recordings in Ableton Live, his music-production software. One technique involved loading sustained sounds made from his voice into Granulator III, an instrument that works with tiny fragments of recorded audio. He varied their playback and recorded the results as new audio.

Recording that processed result as new audio is called resampling. You can then cut up or stretch the new recording, including the effects you’ve already added.

Give the recording a different job

If you record a vocal, it’s easy to judge it as a vocal. Is it in tune? Does the phrase work? Would you want someone to hear you singing it?

Those are reasonable questions when the performance is supposed to carry the song. They become less useful when you’re looking for the sharp beginning of a sound to turn into percussion, or a sustained vowel to stretch beneath the chords.

In Dry Before Wet, I wrote about keeping a lead vocal close and clear before adding effects such as reverb. That advice applies when you want the listener to focus on the singer. When you’re turning a voice into another instrument, the recording doesn’t have to work as an exposed vocal first. Effects may be how you discover what to use it for.

Record yourself holding an “ah” sound. The note may wander slightly higher or lower as more breath becomes audible. Stretching the recording makes those changes unfold more slowly. That can become a sustained background sound, often called a pad. Or cut a short piece from the beginning and repeat it as percussion. Both parts start with the same recording, but use different pieces and different processing.

You can try this in your DAW (I used Ableton Live.) Its warping tools let you change a recording’s timing. Depending on the mode, you can stretch a sound without changing its pitch or change timing and pitch together. Heavy stretching can also alter the sound’s character. Listen before assuming the setting that preserves the original sound most closely is the one you want.

One recording, several parts

In my track Anamorph, I used one 22.7-second vocal recording to create seven parts. My production notes list a pitch-adjusted version, a longer stretched layer and isolated breath sounds. They also include a sustained sound made with granular processing, five vocal layers forming a choir, backward-playing reverb tails that swell in volume and short fragments arranged into a rhythm. The original recording never appears as a conventional sung phrase.

Granular processing plays tiny fragments of a recording, often repeating or overlapping them to create a different sound. Those fragments can become a continuous tone or texture. Whether that result belongs in the track depends on how it sounds alongside the other parts.

One recording can do different things in an arrangement. Its breath sounds might play quietly beneath the main melody, while short fragments from elsewhere in the recording add rhythmic accents. The parts come from the same recording, but contribute differently. Anamorph used seven parts. Another track might only need one.

Saxophonist and producer Laura Misch describes a related process in her June interview with Ableton. She stretches and repeats recorded saxophone phrases, then layers them with effects. Notes can continue after she stops blowing, and overlapping phrases can create chords. She can then build another part in response to what she hears.

In The System Is the Style, I wrote about deliberately limiting your tools and options so you can get on with making music. Here, you can keep the tools you already know and give them a recording you haven’t worked with before. You’re changing what you process, rather than how many tools you use.

Keep the roughness that does something

An uneven recording isn’t automatically more interesting than a clean one. A note that wanders slightly in pitch can make a sustained sound feel less static. It can also sound out of tune with the bass. A breath can make a sound enter gently, or become an irritating hiss after you’ve stretched it and turned it up.

The decision depends on the part you’re making. If the sound needs to hold a stable note, tuning may help. If you’re using a brief noisy fragment as percussion, the original sung pitch may barely matter.

Keep an untouched copy so you can compare each edit with the original. Change one thing at a time: shorten the breath or reduce how much the note wanders in pitch. You could also start playback just after the sound’s sharp beginning. Listen for what each edit removes as well as what it fixes.

Don’t turn this into another obligation to make everything from scratch. If a preset gives you the part the track needs, use it. Recording your own material is useful when it produces an idea worth following. Spending an hour making an ordinary pad doesn’t become a better use of time because the pad began in your mouth.

Try it before the next preset search

Record about twenty seconds of your own voice. Hold an “ah” or “oo,” let the sound change naturally and leave a little breath at the end. You don’t need lyrics or a finished melody. Keep the input meter below its clipping limit so the recording doesn’t distort from being too loud.

Save the original. Work on copies and try these three uses:

1. A short rhythmic sound. Cut out a brief piece with a distinct beginning and repeat it alongside a simple drum pattern. Try making it shorter or placing it between drum hits. Listen for whether it adds a rhythmic accent you want to keep. If the cut creates clicks, use a very short fade at the start or end to soften the abrupt edge.

2. A sustained background sound. Stretch a section where you’re holding a note, then make its volume rise gradually at the start. Play it quietly alongside a chord or bass note. Listen for whether its changing pitch adds movement you like or clashes with the other notes.

3. A moving pulse. Repeat a longer fragment, then vary its volume over time. Leave space between the repetitions. Listen for a rhythm you want to build around.

These are starting points. You don’t have to keep all three. If one doesn’t work, try a different piece of the recording before adding more effects.

Put the most promising sound into a short section of a track. Duplicate it and change one detail, such as shortening a breath that’s become too loud. Adjust the volumes until neither version sounds obviously louder. Play one version at a time with the same backing parts. This helps you hear what the edit changed without giving the louder version an advantage.

Notice whether working with the new sound leads you to move a drum hit, shorten a note or bring a chord in earlier. Those changes may be worth keeping even if you later remove the sound that suggested them. The experiment still helped you write the track.

Before deleting the next awkward take, give a copy one different musical job and hear whether it earns a place.

What part of your own recording would you normally remove before finding out what else it could become?

Leave a comment

Resources

• Ital Tek: Chasing Accidents - Ableton, September 2, 2026. Myson’s account of recording and transforming his source material.

• Mind Abandon - Planet Mu. Release date and the acoustic-guitar source of the kick drums on “Kill Switch.”

• Laura Misch: Of Microphones, Stones & Saxophones - Ableton, June 17, 2026. Acoustic recordings developed through electronic processing.

• Audio Clips, Tempo, and Warping - Ableton’s Live 11 reference manual. Timing, pitch and stretching behavior.

Share

• AlphaTheta says a PRO DJ LINK vulnerability could let someone with unauthorized network access view data on connected computers or USB/SD media.

• The company hasn’t confirmed damage and is withholding technical details while it prepares a fix.

• The lesson is bigger than this bug: if your players share files and state, your booth is already a network.

The Ethernet cable behind the decks is easy to forget about.

You connect the players, mixer and switch, insert a USB and get on with playing music. PRO DJ LINK handles the invisible part. Tracks can be shared between devices. rekordbox can feed music into compatible players. Beat, key and status information can move around the booth without you thinking much about the network carrying it.

Most of the time, that’s exactly how infrastructure should feel. Boring.

On August 8, AlphaTheta gave DJs a reason to look at it again.

The company disclosed a security vulnerability affecting PRO DJ LINK in rekordbox and certain CDJ/XDJ models. AlphaTheta says someone who gains unauthorized access to the PRO DJ LINK network could potentially view data stored on a connected Windows PC or Mac, or on USB and SD media inserted into linked players.

AlphaTheta says it hasn’t confirmed any cases of damage. It’s also withholding the technical details until a fix is available, so there’s no useful reason to invent a scarier exploit story around what the company hasn’t published.

The practical lesson is already enough: the booth has a network boundary whether we think about it that way or not.

Subscribe now

The Network Behind the Decks

PRO DJ LINK sounds like a feature and the cabling tells you what it really is.

AlphaTheta’s documentation describes a LAN connecting compatible players, mixers and computers running rekordbox. Depending on the setup, music stored on computers, USB drives, SD cards and mobile devices can be shared across multiple players. The same network can carry Beat Sync, key information and device status.

That’s useful because the booth behaves like one connected instrument instead of a pile of separate boxes.

The easy assumption is that the network is mostly carrying timing and performance data. AlphaTheta’s advisory makes the storage side visible too. If an unauthorized party gets onto the PRO DJ LINK network, the company says data on the connected computer or performance media may be viewable.

So the question is no longer only, “Did I export the playlist?”

It’s also, “What’s this setup connected to?”

Updating Is Not the Same as Closing the Issue

The rekordbox release history needs a careful read.

Version 7.2.17, released July 30, says: “Enhanced LINK EXPORT security.” Version 7.2.18 arrived August 18 with Spotify-related changes plus general bug and stability fixes. Its notes don’t say the PRO DJ LINK vulnerability is fully resolved.

I wouldn’t infer more than that. The July release note doesn’t identify which security issue it addresses, and AlphaTheta’s August 8 advisory still says the company is preparing a fix and will post progress updates.

So the live advisory wins over guesses about what a version number might mean.

Update rekordbox anyway. Then check AlphaTheta’s current response-status page for the exact hardware and software you use before assuming the issue is closed.

This is normal performance hygiene with a security consequence attached. Current software matters because compatibility matters. Firmware matters because the player has to behave when you need it. Now the network belongs in the same checklist.

Draw the Booth Once

Before your next set, take thirty seconds and draw the network. Literally.

Player 1. Player 2. Mixer. Switch. Laptop. Wi-Fi access point if there’s one. USB and SD media attached to anything on that network.

You don’t need IP addresses or a security diagram. You’re answering one question: what becomes reachable once something joins this setup?

Then do the easy checks AlphaTheta is already recommending. Keep rekordbox current. Don’t carry sensitive files on the USB or SD media you use with PRO DJ LINK. If the setup uses Wi-Fi, make sure it’s secured and password-protected.

I would add one more practical check: know whether the booth switch or access point is only serving the DJ gear or is tied into something broader at the venue. A closed booth network and a shared network aren’t the same thing.

Don’t turn that into paranoia. AlphaTheta’s advisory specifically requires unauthorized access to the PRO DJ LINK network before the described data exposure becomes possible. The connectivity itself isn’t the problem. It’s why PRO DJ LINK is useful.

The useful part is knowing where that connectivity ends.

Share

Keep the Network Boring

DJs have always carried infrastructure with them: backup USBs, cables, adapters, power supplies. The digital booth just moved more of it out of sight.

PRO DJ LINK lets players share music and state because the booth is connected. That connection is part of the performance path now, even when nothing is wrong with it.

The August advisory doesn’t suddenly make networked DJing unsafe. It makes one assumption visible: the setup works best when the network is trusted.

That’s the part worth keeping after AlphaTheta ships the final fix.

Know what’s connected. Keep the software and firmware current. Keep unrelated private data off performance media. Secure the wireless side if you use it.

Then go back to forgetting about the Ethernet cable.

If you drew your current booth as a network instead of a gear list, what would be connected that you normally never think about?

Leave a comment

Resources

• AlphaTheta, Important Notice: Security Vulnerability in PRO DJ LINK, August 8, 2026.

• AlphaTheta, PRO DJ LINK Vulnerability Response Status, live model/software response page linked from the official advisory.

• rekordbox, Release Notes, including version 7.2.17’s “Enhanced LINK EXPORT security” note and version 7.2.18.

• AlphaTheta Help Center, What is the Pro DJ Link function? And what link functions are there?.

• AlphaTheta Help Center, What features can I use when the unit is connected to the PRO DJ LINK network?.

Source note: AlphaTheta’s response-status page is publicly available and lists each affected model. Most affected players show a status of “fix in progress.” This article doesn’t reproduce the full table, out of respect for AlphaTheta’s own documentation. Check the live page for the exact hardware and software you use. This article stays inside claims available on AlphaTheta’s public advisory and support pages.

Analytical note: the argument that network hygiene is part of performance reliability is my interpretation of how PRO DJ LINK combines file sharing, performance state and connected devices. AlphaTheta has not reported confirmed damage from this vulnerability and has not published the technical details.

• Beatport now withholds fully or majority AI-generated tracks while accepting AI-assisted music when the finished track remains majority human-made.

• Traxsource can ask for evidence of human creative input when an AI label is challenged. Its policy specifically mentions screenshots of session files.

• The bigger shift is provenance: how a record was made can now affect ingestion, curation, chart eligibility and discovery.

Subscribe now

A session file used to have one job: let you reopen the track.

Traxsource has given it another.

Its current AI Policy says an artist, remixer, label or distributor challenging an AI label can submit evidence of human creative input. The example Traxsource gives is unusually specific: screenshots of session files.

That doesn’t make an Ableton project some universal certificate of humanity. A screenshot can’t tell the complete story of who made a record or how every decision happened.

But once a music store says the session can matter in a classification dispute, the production archive has changed.

It used to document how you made the record. Now it may also help document that you made it.

Beatport draws the line before the track gets in

On August 12, Beatport expanded its work with Beatdapp to identify AI-generated music during ingestion. Under the updated policy, fully or majority AI-generated tracks aren’t permitted on the platform. They can be withheld during ingestion, with rightsholders notified directly. (Beatport)

AI-assisted music is treated differently. Beatport says it remains eligible when the finished track is majority human-made. Those releases are tagged during ingestion so its curation team knows what it is reviewing and featuring.

That’s more useful than flattening the story into “Beatport bans AI.”

The platform is separating AI used inside a human creative process from a system generating most or all of the finished work. Then it attaches different outcomes to those categories.

One can enter the catalog with additional information attached. The other may not enter at all.

Resident Advisor picked up the change on August 17, and Beatportal still lists the announcement in its Trending Now section as of August 24. This isn’t an abstract argument outside electronic music. A store DJs use every day has turned the distinction into an ingestion rule. (Resident Advisor)

Share

Traxsource can ask to see the session

Traxsource goes further because its policy describes what can happen when the classification itself is disputed.

The store says it can scan music using third-party systems before release and label it substantially as Human or AI Assisted. It also says those systems can make errors. If an account holder challenges the label, Traxsource can consider additional evidence before making its final determination. (Traxsource AI Policy)

Then comes the part I keep coming back to: the session file.

Evidence of human creative input can include material beyond prompting an AI model, such as screenshots of session files or other similar evidence.

Traxsource says evidence can include material showing human creative input beyond prompting an AI model, “such as screenshots of session files or other similar evidence.” The same language appears in Section 14 of its Terms of Service.

The policy doesn’t say every producer needs to submit a DAW project. It doesn’t say a screenshot proves authorship by itself. It doesn’t even claim the scanner is perfect. In February, Traxsource publicly argued the opposite, saying accurate AI detection still has serious limits and that hybrid workflows make the problem harder. (Traxsource position on AI music)

That makes the evidence rule more interesting, not less.

If automated classification can be wrong, the question becomes what a human can produce when the label is challenged.

A project file. Arrangement history. Premasters. Stems. MIDI. Automation. Dated exports. The ordinary debris of making a record starts to have evidentiary value.

Traxsource also attaches consequences to the result. Music it determines is entirely AI-generated can be excluded from its charts or removed from the service. Repeated submissions can put the account at risk.

The creative process used to disappear behind the master. Now part of it may need to remain recoverable.

Spotify separates the artist from the process

Spotify added a different layer on August 11 when it announced AI Persona badges.

The distinction here matters. Spotify says the badge is about the artist’s public identity, not about how the music was made. Artists can self-disclose now. Starting in mid-September, Spotify says listeners will begin seeing the badge on some profiles, in Search and on track rows across playlists. (Spotify)

By default, Spotify says AI Personas won’t be included in editorial or algorithmic recommendations unless a listener has chosen to follow them, for example.

Then Spotify points somewhere else for the creation question: AI Credits and SongDNA. AI Credits let artists disclose how AI was used during music creation. SongDNA gives listeners more context about the artists and contributors behind a track.

So Spotify is separating two questions that are easy to blur together:

Who is the artist?

How was the music made?

Different information. Different product behavior.

Beatport is using AI-related information during ingestion and curation. Traxsource can use it for labeling, charts and removal. Spotify is exposing identity and creation information through separate features, then changing recommendation behavior for one of those identity categories.

Some of this information is public. Some of it sits inside platform operations. Either way, the production process is no longer completely invisible once the master leaves the studio.

Do the six-month test

Open a released project from six months ago.

Now imagine a platform classifies that track as AI Assisted and you believe it’s wrong. Don’t argue the philosophy of it. Ask a much more boring question:

What could you show them?

Maybe you still have the final project with the arrangement intact. Maybe you have dated premasters and stems. Maybe the vocal comp, automation, MIDI, synth patches or dozens of tiny edits tell the story without much explanation.

Or maybe all you kept was the master.

This isn’t an argument for building a forensic evidence package every time you open the DAW. Producers already have enough administrative work.

But preserving the production trail costs much less than reconstructing one later.

Keep the final session. Keep the bounces that matter. Keep accurate credits. If an AI tool materially participated in the work, make a simple note of what it did instead of trusting yourself to remember a year later.

Don’t manufacture evidence after the fact. Preserve the evidence your normal creative process already produces.

And don’t make your music worse trying to perform “human-made” for a scanner. Beatport explicitly allows AI-assisted work when the finished track remains majority human-made. Traxsource describes AI as a legitimate production tool when it sits inside a larger human-driven creative process.

The goal isn’t purity. It’s being able to describe what actually happened.

The session has a second job

Release metadata used to describe the finished object: artist, title, label, writers, genre, ISRC, release date.

Now another question is entering the machinery around the record:

How was this made?

There isn’t one common answer, one shared classification system or one standard of proof across these platforms. Beatport, Traxsource and Spotify are doing different things. That difference matters.

But they’re all attaching information or consequences to some part of the creation story.

That’s the shift I’d pay attention to.

Your DAW session doesn’t need to become a courtroom exhibit. It does need to stop looking like disposable clutter the moment the master is delivered.

Keep the session like it may someday have to tell the story without you standing there to explain it.

If one of your released tracks were questioned six months from now, what would you still have besides the master?

Leave a comment

Resources

• Beatport, Beatport Expands Beatdapp Partnership to Detect AI-Generated Music, August 12, 2026.

• Traxsource, AI Policy, current v1.1.

• Traxsource, Terms of Service, Section 14: AI Policy.

• Traxsource, Our Position on AI Music, effective February 19, 2026.

• Spotify, Introducing a New Label for AI-Generated Artist Identities on Spotify, August 11, 2026.

• Resident Advisor, Beatport expands Beatdapp partnership to detect AI-generated music, August 17, 2026.

Analytical note: the argument that production provenance is becoming operational metadata is my inference from these policies and product features. The platforms use different classifications and consequences. I am not claiming they share one industry standard, that AI detection is infallible or that a DAW screenshot alone proves human authorship.

• Ableton Live 12 Suite (version 12.3 or later) can split a finished track into Vocals, Drums, Bass and Others, then warp, chop or recombine the results.

• Engine DJ 5.0 can render four-part stems directly on RANE SYSTEM ONE hardware without a computer while normal playback continues.

• Those stems are estimates, not your original multitracks. Ableton says sounds can land in unexpected categories. Producers should hear what their master becomes when software pulls it apart.

Subscribe now

A finished master still leaves your studio as one stereo file. The software waiting on the other end increasingly treats that file as raw material.

Ableton Live 12 Suite can split recorded audio into Vocals, Drums, Bass and Others with a source-separation model that runs locally. Engine DJ 5.0 can render four-part stems directly on a RANE SYSTEM ONE without a computer. In April, Beatportal compared stem separation across rekordbox, Serato DJ Pro, djay Pro, Traktor Pro, VirtualDJ and Engine DJ.

The interesting producer problem isn’t which separator wins a shootout. A DJ can now get playable pieces from a master they only ever heard as one finished track.

Producing music, you know that the pieces aren’t your session stems. They’re the machine’s simplified estimate of them.

The master can be turned back into parts

Ableton shipped built-in Stem Separation with Live 12.3 Suite on November 25, 2025. Its current manual says the resulting clips behave like other audio in Live: they can be warped, chopped, rearranged and processed. Ableton lists remixing, mashups, sampling and DJ mixes among the intended uses.

Then Live 12.4 made the workflow more surgical. Since May 5, 2026, you can separate only a selected section of an Arrangement clip. You can also choose two or three separated parts and merge them into a new track. Ableton’s own example is removing the vocal and merging Drums, Bass and Others into an instrumental. The details are in Ableton’s 12.4 notes.

That changes the amount of preparation needed for an edit. You don’t have to pull an entire song apart before using one section. Sixteen bars can be enough.

Engine DJ moved the same basic idea onto standalone performance hardware. Engine DJ’s 5.0 announcement says RANE SYSTEM ONE can render four-part stems on the device with no computer attached. The track can keep playing while the render happens in the background.

There’s an important limit. This isn’t instant separation the moment you touch a pad. Engine’s support documentation says the stem controls become available after rendering finishes. If that track is already loaded, the controls are ready as soon as the render completes.

In other words, a DJ can arrive with one finished file and leave the separation step to the software or, on some hardware, the deck itself.

The machine doesn’t recover your session

This is the part I’d pay attention to as a producer.

Ableton says its model analyzes the spectral and temporal characteristics of the mixed signal. It isn’t opening your project and finding the tracks you named Kick, Bass, Vocal, Pads and Synth. It’s estimating components from the audio you already summed together.

Ableton is unusually direct about what can happen next. The manual says high frequencies from hi-hats may appear in Others. A synth with chorus can show up in Vocals. Material can land in a stem you didn’t expect because the model is classifying what it hears, not reading an arrangement.

That means the boundaries in the separated version can cut across boundaries you drew in the session.

A sound you thought of as part of the percussion may partly live somewhere else. A texture that only made sense glued into the full mix may become obvious when the drums disappear. The separated version can reveal relationships that were never created as solo objects because they were never supposed to be solo objects.

That isn’t automatically a defect. A DJ may only need a few bars of usable material under another track. Ableton itself frames stem separation as material for remixing, mashups and DJ mixing. The isolated part doesn’t have to sound like the original multitrack to become useful.

Don’t mix for the separator

I wouldn’t start moving sounds around because a separation model puts a synth in the wrong bucket.

The master still has to work as a track. Stem separation is a check on what someone else can extract from it, not a new mix target.

Ableton even offers High Speed and High Quality modes, and its manual suggests trying both when the results aren’t what you want. That alone should kill the idea that there’s one fixed, objective set of hidden stems waiting inside every stereo file. The output depends on the separation process and the source material.

We already change listening conditions to learn something about a mix. Stem separation is different because it doesn’t only change playback. It changes the material the listener gets.

Maybe the answer is ugly. Fine. Your track wasn’t written to be four independent solo tracks. But if one of the separated parts contains a transition, texture or rhythmic fragment a DJ could actually use, that’s worth knowing too.

Run 16 bars through it

Take the released master, because that’s the file another DJ is likely to have. Pick 16 bars where the arrangement is doing real work. A transition is useful. So is a section where the full groove is in.

Run that section through Live 12.4 in High Quality mode. Listen to the full master once. Then solo Drums, Bass, Vocals and Others one at a time. Don’t fix anything yet.

Write down what surprises you. Which sound landed somewhere you wouldn’t have put it? Is one separated part more useful than you expected? Does removing a stem uncover a movement that was hard to notice in the full track? Is there an artifact bad enough that the part falls apart by itself?

Then make four versions with one category missing each time. Live 12.4 can merge selected stems into one new track, so this is quick. Play those versions like you were building an eight or sixteen bar transition into another track.

Don’t ask whether the stems are perfectly clean. Ask what a DJ can actually do with them.

What the DJ gets is worth hearing

The test may not change a single mix decision. That’s okay.

If the master works and the separated parts are messy, the separator may simply be making a rough estimate from a dense stereo file. If a strange piece comes out that works under another track, you’ve learned something different: your finished track contains performance material you never bounced on purpose.

The useful outcome is knowing what comes out. Sometimes the result will be clean enough to use. Sometimes it will sound mangled. Either way, you’ve heard a version of the track that current production and DJ tools can create without your session.

Before the next release, I’d run the master through stem separation once and listen without trying to repair it. I’d want to know what the software hands to the person on the other side.

Which 16 bars of your latest track become more interesting when one of the four parts disappears?

Share

Resources

• Ableton, Stem Separation - Live 12 Manual. Current documentation for the local deep-learning separation model, four stem types, creative uses, editing behavior and examples of material appearing in unexpected stems.

• Ableton, Live 12.3 is here - see what’s new, November 25, 2025. Confirms the public release of built-in Stem Separation in Live 12.3 Suite.

• Ableton, What’s new in Live 12.4, May 2026. Documents time-selection separation and Merge to Single Track.

• Engine DJ, Introducing Engine DJ 5.0 - On-Board Stems Rendering for the RANE SYSTEM ONE, May 14, 2026. Documents four-part on-device stem rendering without a computer and background rendering during playback.

• Engine DJ, RANE SYSTEM ONE - On Device Stems Rendering. Clarifies that stem functions become available after rendering completes and that playback can continue while another track is rendering.

• Beatportal, Who Does Stem Separation for DJs Best? All DJ Platforms Compared (2026 Edition), April 16, 2026. Industry comparison covering rekordbox, Serato DJ Pro, djay Pro, Traktor Pro, VirtualDJ and Engine DJ, plus studio-focused separation tools.

Analytical note: the argument that the finished master is becoming easier to treat as performance material is my inference from these product capabilities. None of the sources measures how often DJs actually separate a given track in public. I’m not claiming that every DJ does this or that producers should mix for separation algorithms.

• Spotify’s Talk to Spotify beta lets eligible Premium listeners ask for music in ordinary language and refine the request conversationally.

• Explicit genre and tempo prompts landed pretty well in my test. Once I removed the genre labels and described the sound in producer language, the results wandered.

• Giving Spotify ZenOne as an artist anchor brought the first results back toward a much more credible hypnotic-techno neighborhood.

There’s a particular kind of track that’s annoying to describe. It has enough psychedelic movement that calling it techno feels incomplete, but enough restraint that psytrance doesn’t quite describe it either. Then an upload form asks for a genre and something complicated becomes a dropdown choice.

Genres are useful. DJs need crates, stores need organization and listeners need somewhere to start. But the label is still an approximation. Spotify just gave me another way to ask for the music: a sentence.

The beta was sitting right there

Spotify announced Talk to Spotify on July 14, 2026. It’s rolling out gradually in English to Premium listeners 18 and older in the U.S., Ireland and Sweden on iOS and Android. Listeners with access can type or speak from Home or Now Playing, then refine what they hear with follow-up requests.

In my app, it appeared on Home as a field that says “Ask for some music.”

Tap it and Spotify puts the caveat right up front: you’re about to talk to AI, it can make mistakes and you shouldn’t share sensitive data.

Spotify explains what the interface can do, but not how every descriptive word gets mapped to a recording. I can test the input and inspect the output. I can’t see most of the machinery between them.

The easy words worked

I started with “Play hypnotic techno around 140 BPM.” Spotify opened with Yan Cook’s “Worm”, which Beatport lists as Techno (Peak Time / Driving) at 138 BPM. Pretty close on the parts I could verify.

Then I added “with psychedelic movement” and got Amotik’s “Ikat.” My third prompt asked for “fast, hypnotic electronic music around 140 BPM with psychedelic movement, driving bass and very little melodic content.” Spotify led with Yan Cook’s “Blades”, which Beatport lists at exactly 140 BPM under Techno (Raw / Deep / Hypnotic).

That isn’t proof Spotify understood every adjective. Tempo and established genre language are easier to verify from the outside than “psychedelic movement.” The early results still made enough sense that I kept pushing.

Then I took away the labels

The test I cared about most removed techno, psytrance and Psy-Techno entirely. I asked for “fast electronic music with driving low end, psychedelic modulation, sparse melodic information, hypnotic repetition and long tension without giant festival-style releases.”

Spotify opened with Big Wild’s “Strangeland.” Beatport lists it as House at 140 BPM. Farther down the queue were Galantis and Ian Asher’s “Runaway (U & I)” and Ian Asher with Olly Alexander’s “Desire,” which sounded like a pretty different neighborhood from the one I thought I’d described.

Beatport’s categories aren’t a final truth about sound, but they’re useful here as an external reference. In this run, the tempo neighborhood survived while the aesthetic neighborhood became much harder for me to recognize.

Psy-Techno didn’t rescue it

Beatport introduced Psy-Techno under Techno (Peak Time / Driving) on January 27, 2026, describing it as a crossover between techno’s hypnotic pulse and psychedelic sound design, driving basslines and Psy-Trance-style layering.

So I added: “Something in the Psy-Techno area.” Spotify renamed the request “Psy-techno music request” and opened with Aaron Hibell’s “destroyer of worlds”. Beatport lists it as Dance / Pop at 83 BPM. One track doesn’t settle anything. Still, the label I expected to help didn’t move the first result where I expected it to go.

The conversation itself also hit a beta edge. “Keep the psychedelic movement, but less psytrance and less hard techno” returned “Something went wrong. Please try again.” I retried five times, then “Make it more hypnotic” failed too. Starting a new conversation fixed it for me. I don’t know what broke.

Then I said ZenOne

For the last test I gave Spotify an artist anchor: “how about songs by artists like ZenOne. then play artists I probably haven’t heard before that fit this sound.”

Spotify answered: “Going ZenOne-adjacent and keeping it new-to-you. Same psy-techno tension, just more unfamiliar artists.” The first result was Gary Beck’s “Ghost”, which Beatport lists at 140 BPM as Techno (Raw / Deep / Hypnotic). Next came DJ Dextro’s “1000 Places”, also 140 BPM in the same category.

That was noticeably closer to the territory I expected. And yes, my first reaction was still: damn, where’s ZenOne? Then I reread my prompt. I’d asked for artists like ZenOne and for artists I probably hadn’t heard before. Not playing ZenOne itself was reasonable.

In this one personalized session, my descriptor-only prompt wandered away from the neighborhood I had in mind. Adding ZenOne produced first results that Beatport categorizes as Raw / Deep / Hypnotic Techno at 140 BPM. I can’t tell you why Spotify made those choices. Its public explanation doesn’t expose enough of the recommendation process for me to pretend otherwise.

A sentence is another handle

I started with the idea that conversational discovery might make a producer’s sonic fingerprint easier to find. After using it, that feels too clean. Spotify handled some explicit constraints well in my session, then became much less convincing to me as I shifted toward compound descriptions of movement, restraint and tension.

My read after one session is that natural language works better as another steering layer than as a replacement for genre or artist relationships. Beatport is making one crossover taxonomy more precise while Spotify lets listeners say more than the taxonomy can hold.

There’s a producer experiment in that gap: describe one of your tracks without the artist name or genre. Use pace, density, movement, tension, space and whatever slightly strange words you’d use with another producer. Put that sentence into Spotify, then run it again with your artist name as the anchor. Compare where each one goes.

Sources & Further Reading

• Spotify. “Just Say the Word: A More Personal Way to Ask, Discover, and Listen.” July 14, 2026. https://newsroom.spotify.com/2026-07-14/talk-to-spotify-announcement-beta/

• Spotify Support. “Talk to Spotify to discover and shape your listening.” https://support.spotify.com/us/article/talk-to-spotify/

• Spotify. “Prompted Playlist in Beta Coming to Premium Listeners in More Markets.” January 22, 2026. https://newsroom.spotify.com/2026-01-22/prompted-playlists-expansion/

• Spotify. “Spotify’s Discovery-Driven Playlists Keep Your Weekly Music Rotation Feeling Fresh.” July 10, 2026. https://newsroom.spotify.com/2026-07-10/discovery-playlists-release-radar-control-updates/

• Beatport. “Beatport Launches New Sub-Genre: Psy-Techno.” January 27, 2026. https://www.beatportal.com/articles/1257068-beatport-launches-new-sub-genre-psy-techno

A VST3 isn’t inert media. It’s executable code your DAW can load, which makes every plugin you can’t trace back to a trusted source worth another look.

I’m in Vegas this week for Black Hat and DEF CON, surrounded by people whose job is basically to look at normal software and ask what else it can do. At some point that question crossed over into music production for me, and I started thinking about what’s actually sitting in people’s plugin folders.

Open the folder your DAW scans for plugins.

After enough years making music, it can turn into a kind of sediment. Old installs, utilities you tried once, versions you forgot you still had. Names that make perfect sense until you try to remember exactly where the file came from.

That matters because a plugin isn’t a preset or a sample your DAW just reads. It’s code.

What the format actually does

Steinberg’s VST3 specification describes a plugin module as executable software. On Windows the module contains a DLL. On macOS it’s a bundle containing a Mach-O binary. A host such as Ableton Live discovers those modules, initializes them and loads their components when it needs them.

That sounds obvious when you say it out loud. It doesn’t always feel obvious when you’re staring at a folder full of compressors.

In February 2024, a security researcher publishing as infosecnoodle demonstrated the distinction using Steinberg’s own Hello World VST3 example on Windows. The researcher added a system call, rebuilt the plugin and placed it in the standard VST3 location. Opening the modified plugin in Ableton executed the added command.

The researcher also showed that when Ableton detected a new or modified VST in the plugin directory, plugin initialization could execute code during the scan process. The execution appeared under Ableton Index. For a persistence demonstration, the researcher went further and injected code into AbletonPushCpl.exe, an Ableton-related process that could remain running after the DAW closed.

This wasn’t a vulnerability in Ableton or VST3. It was a proof of concept showing what executable plugin code is capable of doing when you choose to load it.

That’s the part I keep coming back to.

The plugin folder isn’t passive storage. It contains programs your music software may execute.

The old version only wanted your CPU

None of this started with 2026.

In June 2019, ESET documented LoudMiner, a cryptocurrency miner distributed inside pirated audio software for Windows and macOS. ESET traced the campaign back to at least August 2018. One website alone listed 137 VST-related applications, with lures carrying names producers would recognize immediately: Ableton Live, Reason, Sylenth1, Nexus, Reaktor and Auto-Tune.

LoudMiner’s architecture was almost absurdly elaborate for what it wanted. The malware ran mining software inside a virtualized Linux environment, using QEMU on macOS and VirtualBox on Windows.

All that machinery to steal compute.

The 2026 version of this story gets more intimate.

In February, security company Iru published an analysis of a mass-distributed macOS loader disguised predominantly as cracked music software. Researchers found the same hidden configuration inside more than 100 malicious DMG files. One analyzed request actually carried this lure in its parameters:

Download iZotope Ozone Pro ... Plugin Crack

The name isn’t evidence that iZotope itself had anything to do with the campaign. Quite the opposite. Its product name was being borrowed as bait.

The chain used a Bash script to retrieve another stage and included a ClickFix-style prompt designed to convince the victim to paste a malicious command into Terminal. Iru observed payloads including MacSyncStealer and Odyssey.

One detail makes the scale of the thing easier to see. The delivery chain contained an affId parameter. Iru identified it as an affiliate identifier and concluded the campaign was operating as part of a pay-per-install network.

Someone wasn’t just making malware. There was distribution economics behind it.

CloudSEK separately analyzed MacSync, and its findings are the part producers should probably care about more than the malware’s name. The stealer targeted browser credentials and cookies, macOS keychains, cryptocurrency wallets, SSH keys, AWS credentials and Kubernetes configuration files. It also searched Desktop, Documents and Downloads for selected file types.

Then it compressed what it collected, uploaded the archive and deleted the local archive after exfiltration.

That’s a very different proposition from somebody borrowing your CPU to mine Monero.

The thing actually worth stealing

For a producer, the plugin may not even be the most valuable thing on the computer.

The unfinished work is.

Late last month, Ariana Grande filed a lawsuit in Los Angeles County Superior Court against unidentified defendants she alleges spent years obtaining unreleased music, photos and other private material through people in her professional circle.

The allegations matter here because of the entry points. According to the complaint, credentials for a photographer’s Dropbox account were obtained in 2019. In 2020, the mobile device of a producer who had worked with Grande was allegedly hacked. The complaint says still-in-production masters, demos and recording-session footage were taken.

Then came phishing.

In January and February 2024, the complaint alleges that attackers created a Gmail account and a lookalike domain impersonating a photographer, then used them to convince that photographer’s digital technician to hand over private material. Grande alleges that 45 unreleased songs were hacked, stolen and leaked in 2023 alone. The complaint also says stolen material was sold using services including PayPal and Cash App before purchasers redistributed it.

Those are allegations. The defendants haven’t been identified and the claims haven’t been adjudicated.

Nothing in the complaint says a malicious music plugin was involved. These are not the same attack.

What connects them is the topology.

In the incidents Grande describes, the access points weren’t some imaginary vault labeled ARIANA GRANDE MASTERS. They were collaborators. A photographer. A producer. A technician. The people already close enough to the work to have it.

Anyone who has ever received stems before release, held a vocal under NDA or opened someone else’s unfinished project should recognize the shape of that immediately.

Your studio can contain somebody else’s secrets too.

Grande had already put her frustration with leaks rather less clinically in 2024. Talking with Zach Sang about leaked material from her sessions, she addressed the people responsible:

“I’ll see you in jail, literally.”

Fair enough.

Piracy isn’t the whole variable

It would be easy to turn all of this into a lecture about cracked plugins. I’m not interested in doing that.

For the record, every plugin and piece of production software I run is paid for in full. That isn’t a flex, and it isn’t a judgment of anyone who has made different choices under real budget pressure. I mention it only because provenance is easier to reason about when you can trace each install back to a receipt and an official download.

There’s a more useful distinction.

On July 29, Tomislav Zlatic released VATRA, the first premium plugin from his Flame Sound label. According to Zlatic, VATRA shipped without copy protection. No activation server, no license manager, no keys.

Roughly two days after launch, an unauthorized copy was already circulating.

There was no protection layer to defeat first.

That’s useful because it separates two things that tend to get collapsed into one argument:

Piracy is a licensing problem. Provenance is a security problem.

Share

I couldn’t find a technical teardown of the unauthorized VATRA copy, so I’m not going to imply it contained anything malicious. There’s no evidence I found that it did.

The point is simpler. Once software leaves the developer’s distribution path, the name on the file isn’t proof of what is inside it.

The same pattern extends far beyond audio software. Kaspersky documented the Stealka infostealer in late 2025 spreading through game mods, cheats and software cracks. One sample masqueraded as a Microsoft Visio crack. The malware targeted data from 115 browser extensions and 80 cryptocurrency wallet applications.

In February 2026, Kaspersky published research on RenEngine, a loader distributed through pirated games and professional software. Researchers found dozens of websites pushing the loader through cracked software, including CorelDRAW. Earlier RenEngine infections delivered Lumma Stealer. More recent chains delivered ACR Stealer, with Vidar also observed.

Neither campaign specifically targeted music producers.

That’s exactly why I care about them.

The producer looking for one missing plugin at 1 AM isn’t living in a special music-industry threat model. They’re downloading executable software from the same internet everyone else is.

Twenty minutes, not a lecture

I’d normally put a listening experiment here.

Not this time.

Do an inventory instead. Open the locations your DAW scans for plugins and, for each plugin you actually use, answer one question:

Where did this copy come from?

Not whether you paid for it. Where it came from.

The vendor’s installer? An official package manager? A developer you know? An old drive? A download folder from three machines ago? Some archive whose origin you honestly don’t remember?

The ones you can’t source are the list.

Don’t panic. Don’t start deleting random files in the middle of active projects either. Just know what they are.

Then look one layer outward. Where do your unfinished projects live? Where are the stems people have trusted you with? What gets synchronized to cloud storage? Which old backups still exist? Who can reach them?

Finally, check the accounts around the work: your distributor, cloud storage, email and collaboration tools. Unique passwords. MFA or passkeys where they’re available. No reused credentials just because all of this feels less consequential than a bank account.

It isn’t.

Twenty minutes.

Maybe longer if the backup situation has become archaeology.

The quiet scan

There is one uncomfortable ending to this.

You can’t prove a plugin is trustworthy just because an antivirus scan is quiet.

The 2024 VST proof-of-concept author reported low detection rates for the malicious VST samples they tested and argued that many security products appeared not to inspect the format closely. That’s one researcher’s observation, not evidence that every security product misses malicious plugins.

From my side of the fence, though, I’ve seen enough security tooling over the years to be skeptical of the comfort it sells. Some of it is excellent. Some of it is expensive snake oil with a dashboard. And a lot of the weaker stuff can be bypassed more easily than the marketing would have you believe. A clean scan is useful evidence, but it’s not the same thing as knowing the file is safe.

But the larger principle doesn’t need that claim.

A clean scan is evidence. It isn’t provenance.

Knowing where the software came from is a different question.

So the inventory isn’t a magic fix. It’s just the part you can actually see.

A plugin solved a problem in the mix, so you kept it. Six months later it became part of the furniture. Three years later nobody remembers where the installer came from.

That’s normal studio entropy.

It also happens to be an attack surface.

Everything in that folder doesn’t run all the time. But everything executable in there can run when the host loads it.

That’s enough reason to know what you put there.

Resources

• VST 3 Plug-in Format Structure - Steinberg’s official documentation for the VST3 bundle and executable-module structure on Windows and macOS.

• VST Audio Plug-ins for Initial Access and Persistence - infosecnoodle, February 17, 2024. The Ableton/VST3 proof of concept covering command execution, plugin initialization and the AbletonPushCpl.exe persistence demonstration.

• LoudMiner: Cross-platform mining in cracked VST software - ESET Research, June 20, 2019. Documents LoudMiner, the 137 VST-related applications found on one distribution site and its QEMU/VirtualBox architecture.

• macOS Malware Analysis: Music Plugin DMG Loader - Iru, February 19, 2026. Primary analysis of the 100-plus malicious music-software DMGs, Ozone-themed lure, ClickFix component, MacSyncStealer/Odyssey delivery and pay-per-install affiliate identifier.

• MacSync Stealer: SEO Poisoning and ClickFix-Based macOS Malware Delivery Chain - CloudSEK, March 18, 2026. Detailed analysis of MacSync’s credential, keychain, SSH, cloud-configuration and document theft plus its exfiltration process.

• Ariana Grande sues alleged hackers over unreleased music leaks - ABC News, July 27, 2026. Reporting based on the complaint filed in Los Angeles County Superior Court, including the alleged 2019, 2020 and 2024 intrusion timeline.

• Ariana Grande Calls Out Hackers After Unreleased Songs Leaked Online - People, February 27, 2024. Source for Grande’s Zach Sang Show remarks about the leaked material.

• My first plugin was pirated in 48 hours. Here’s why it still has no copy protection. - Tomislav Zlatic, August 5, 2026. First-person account of VATRA’s no-copy-protection release and the unauthorized copy appearing roughly two days later.

• The Stealka stealer hijacks accounts and steals crypto while masquerading as pirated software - Kaspersky, December 18, 2025. Documents Stealka’s distribution through cracks, cheats and mods and its targeting of browser extensions and cryptocurrency wallets.

• Kaspersky identifies RenEngine loader distributed through pirated games and software - Kaspersky, February 23, 2026. Documents RenEngine distribution through dozens of pirated-software sites and the Lumma, ACR Stealer and Vidar payloads.

You run the mix through the AI chain. Ozone’s correlation meter throws a flag on the low end: out of phase, mono compatibility at risk. You nudge the sub. The meter reads clean, you bounce, done.

The next day the bass still feels loose.

You don’t know why. But you’re not surprised, either. The meter wasn’t lying: iZotope’s own Ozone documentation is explicit that the Imager flags the visual signature of a phase problem, rays drifting past the 45-degree safe lines and leaves the actual fix to the engineer. The algorithm didn’t find the problem. It told you one existed, in a signal you’d been listening to for weeks without hearing it.

Meridian taught me this one. 135 BPM, multiple drops, more automation curves than I care to count, and by the end I’d played that low end so many times my brain was filling in the version I intended. The pre-release bounce metered clean: correlation steady, no flags, every readout checked out okay. Then I played it loud on the big system and the sub under the second drop wouldn’t hold. Kick fine, weight gone. I blamed the room first. Then the limiter. It took most of a weekend to run out of gear to blame. The hard part of that moment was never the fix. It was admitting I didn’t hear it to begin with.

Sit with what actually happened there. The correlation meter measured the signal, and the signal really was out of phase: no argument, the math is the math. What the meter can’t do is name the cause. The same smeared image on a vectorscope could be a stereo-widened synth or a double-track panned wrong or something else entirely. The module doesn’t care. It draws the shape and hands the question back to you.

And that’s the diagnosis: you’d played that low end a hundred times and your brain filled in the bass it expected instead of the bass that was there. If you can’t feel the looseness, no algorithm can fix it for you: it can only point at the meter and wait.

iZotope markets Ozone 12’s Master Assistant as a “creative co-pilot for engineers of all levels,” and the pitch is that watching the assistant make its moves builds the very ear it’s standing in for.

Watch what it actually does, though. It makes the meters look good. Whether the track still feels like it belongs when you play it in the car the next day is a different question and the assistant can’t answer it: that verdict never shows up in a readout. A co-pilot that does the listening for you isn’t building your ear. It’s standing in for it.

But look at who’s actually leaning on the tools. LANDR commissioned a survey of 1,200-plus musicians last fall (vendor-run, self-selected pool, so salt it accordingly) and 87% said they’d folded AI into some part of their process. The sharper number sits underneath: the survey showed beginners using AI song generators at about twice the rate of pros - 51% versus 25%. The difference isn’t in the tool. It’s in the ear.

Training wheels come off. I don’t see that happening in those numbers. The danger is learning to trust the machine’s hearing before you’ve built your own.

There’s precedent for meters missing what matters. Apple ships a command-line tool called afclip for exactly this reason: it finds the inter-sample peaks an encode to AAC can produce, clipping a standard level meter never shows. The readout can say zero overs and the encoded file can still crack up on a big system. That gap is why the tool exists at all.

Loudness has the same blind spot. Spotify pulls playback to -14 LUFS by default (measured per the ITU-R BS.1770 standard, and Premium listeners can shift that target). Apple’s own Digital Masters brief admits that masters pushed loud will play back quieter under Sound Check, which “can make tracks actually sound weaker.”

An algorithm can hit -14 all day. It can’t tell you the mix has no weight at that volume, because weight isn’t on any meter.

The tool doesn’t ask what you meant. It doesn’t learn what weight means in your low end or why this drop needs to hit harder than the last one. It reflects its training data back at you and whatever you can’t already hear falls straight through the mirror.

So the algorithm is honest about exactly one thing: what it measured. Everything else is still your job.

Ascendance was the other one. Every readout clean on the final bounce: levels, phase, spectrum all sitting where I wanted them. And the drop was polite. Not wrong, just polite. Nothing any meter calls a problem. So I did the only thing left: killed every analyzer on the screen, turned the lights off and played it loud, twice. In the dark it was obvious. The drop wasn’t landing because nothing was getting out of its way: the 2-4k range was carrying too much all the time, so when the moment came there was no room left to arrive in. That’s where the decisions about what occupies that range, the ones I’ve written about since, actually came from. The fix took an evening. Finding it took the dark.

Case in point: there’s a new track I’m working on, Eclipse, looping in my DAW while I write this. Sometimes out loud on the monitors, right now in my earphones, over and over. Not to mix it. Just to sit in it long enough that my ear/mind stops filling in what I meant and starts telling me what’s actually missing. No meter is going to hand me that.

The readouts can all go quiet and green, and something in the drop can still not be there. The machine will never be the one to tell you.

So next time every meter reads green, ask yourself the only question that counts: have you actually heard this mix, or have you just watched it pass? What has your ear caught lately that no readout ever flagged?

If this look at where the tools stop and your ears start was useful, subscribe to ZenOne Music for more deep dives into perceptual mixing, underground production, and the physics behind the sound.

See you in the next one.

Resources

• Loudness normalization on Spotify - Spotify’s own official explanation of its -14 LUFS default and user-selectable loudness tiers.

• Apple Digital Masters: Music as the Artist and Sound Engineer Intended (technical brief) - Apple’s own mastering/encoding spec, including the afclip inter-sample-clipping tool and Sound Check behavior.

• From Editing to Mastering: AI Research Insights at ISMIR 2025 - Sony AI’s account of ITO-Master, built to fix the ‘no way to refine it’ gap in automated mastering.

• MixAssist: An Audio-Language Dataset for Co-Creative AI Assistance in Music Mixing - COLM 2025 paper/dataset on real producer-to-amateur mixing dialogue, framed against automation-only AI tools.

• AES Student Recording Competition - official rules - Confirms the current, explicit ban on AI mixing/mastering services in student competition entries.

• 68th GRAMMY Awards Rules & Guidelines (official rulebook PDF) - Recording Academy’s ‘Generative Artificial Intelligence’ eligibility rule, verbatim, for the Feb 1, 2026 ceremony.

• LANDR Study Uncovers How 87% of Musicians Really Use AI - LANDR’s Nov 2025 survey of 1,200+ musicians on AI adoption across the creative workflow.

• iZotope Ozone 12 Advanced (product page) - Vendor’s own ‘creative co-pilot, not autopilot’ framing of the Master Assistant AI feature.

Leave a comment

I put out a track called Cascade (Original Mix). It was an experiment, and I said so. Then I sent it around to people who listen to new music for a living and got every opinion at once.

One added it and called it a standout. A few said it ran too long. Fair enough, but the long version was meant to unfold as a journey. It asks you to stop waiting for the next obvious cue, let the music fill the space, and follow the patterns as they develop. Some people seem to lock into that internal language immediately. Others don’t, which is part of making something that isn’t built around the shortest route to the payoff. A shorter cut was already finished for release the following week and would eventually sit alongside my other radio edits in The Algorithm’s Cut.

One curator called it an “overloaded cartoon soundtrack,” said it sounded like an unfinished draft, and told me to rebuild it around the instrumentation and arrangement of a reference track or it would “have no form.” Even the samples, loops, synth presets and BPM, apparently, had to stay inside the approved boundaries of a specific genre. There were no samples or loops. Every sound was synthesized and shaped to interact with the others, creating rhythms, tensions and textures that emerge between the layers rather than living inside any one of them. Then, after suggesting I find someone else to finish the mix and master, the same note offered those services for hire.

That stung for a bit. Then I separated the useful question from the genre policing and the sales pitch, and filed the rest where it belonged. Genre labels, arrangement conventions, sound selection and form are arguments about taste and intention. You can lose those arguments and keep walking. I make experimental music on purpose.

But two of them said the bass was weak, and that isn’t purely an aesthetic judgment. It either translates or it doesn’t. So I didn’t argue, and I didn’t trust my own ears either. I listened on multiple systems, then measured it. Before I looked, I decided what would count as losing: if the low end came apart when I summed it, they were right and I had a remix to do.

The part of the low end that isn’t an opinion

Everything in 2026 is pushing you wider. Immersive this, spatial that, a stereo imager on every bus. So when your bass feels small the move seems obvious: make it wider. Drop a widener on the low end and it blooms. Huge.

Up top that’s a gift. The air, the pads, the reverb tails all earn their width.

Down low, width is a trap. A specific, measurable one, which is the good news.

Subscribe now

Why you can’t hear a wide sub, but a mono rig can

You can’t reliably place a low note in space, and the reason matters, because the folk version is wrong. Your ears use two cues: which ear hears it louder, and which hears it first. Down low the first one is gone. A bass wave is many times longer than your head, so it bends around it and both ears get the same level. What’s left is timing, and timing down there is fragile. In a real room the reflections and the modes smear it.

So this isn’t “bass is non-directional.” In a lab, on headphones, with a pure tone, people do better than the folklore claims. In a room, hunting for the sub, they don’t. It’s why one sub in the corner fills a club and nobody looks for it. Which means a stereo sub isn’t buying you width you can hear.

What it is buying you is a phase relationship between left and right. And nearly every system that matters folds the low end back together: club rigs run their subs in mono, phones and Bluetooth pucks are one speaker, vinyl has to be cut with mono bass or the needle jumps the groove. When left and right aren’t the same signal down there, summing them cancels, and your bass thins by a different amount depending on where you’re standing.

The ears lie, the room lies, and this is the third liar in the series: the one you introduce yourself, with a plugin, chasing size.

So you get width you can’t hear, paid for with low end that vanishes on the systems you want to land on. Bad trade. I call the line where it flips the mono floor. Above it, stereo is size. Below it, stereo is risk.

I wrote my own version of this in February, as a rule: keep everything under 150 Hz mono. I’d revise that now, and not because 150 is wrong. Because handing you one number was the mistake.

The floor isn’t a constant. It sits wherever your stereo content stops being audible as width and starts being a bet, and that moves with your bass sound, your arrangement, your genre. Ask a mastering engineer and you’ll hear 150, sometimes 200. February me said 150. Cascade measured nearer 120. We’re all right about our own material, which is the tell that what you want is a method, not a number.

Cutting engineers have folded bass to mono for decades, with a tool named the elliptical equalizer, because vinyl forced them to long before phones did.

Putting my own track on the bench

The weak-bass complaint. Instead of defending Cascade or believing it, I measured.

Below 120 Hz, the left and right channels came back at +0.993 correlation. Meaning they’re already one mono signal. Fold the whole track to mono and the low end moves 0.02 dB. In the drops, where the bass has to hit, it held at 0.999. There was nothing down there to cancel. Whatever “weak” meant, it wasn’t a phase problem. My sub was already mono. (Part of me wanted the bug. A bug you can fix in an afternoon.)

One stretch does wander. Around 3:45 the correlation falls to 0.39 and nearly a third of the low band swings to the sides. It’s a quiet passage for introspection, which is why it happens and why it barely matters: there’s almost no bass in there to lose. I’m flagging it because anyone who runs this will find it, and I’d rather say it first.

You don’t need my toolchain, just a correlation meter. Your DAW has one.

Put it on your master or bass bus and play the drop. Near +1 means the low end is mono and safe. Drifting toward zero, or below it, means stereo energy under the mono floor waiting to cancel on somebody’s system. The mono button still works, but it’s the blunt version. It tells you the bass thinned. The meter tells you where, and at which bar.

The fix is boring and it works. Mono everything below the floor, and find the floor rather than inheriting it: start at 120 Hz, walk it up toward 150 if the meter still wanders, down toward 100 if you’re losing something you can hear. One control in most DAWs. Keep the width up in the harmonics where the ear can use it. Then check the meter holds near +1 where the sub lives.

Share

What the meter is for

An “expert’s” opinion and a correlation meter aren’t the same kind of thing. One is a taste you can take or leave. The other is the room and the club and somebody’s phone, testifying in advance.

Experiment with everything above the mono floor. That part is yours, and nobody gets a vote unless you let them. Below it, stop arguing and let the physics decide.

What’s the last track you widened to make it feel bigger, and would its low end survive a mono club rig tonight?

How I measured this

Python, reading the released 44.1 kHz stereo master with soundfile, filtering and spectra from SciPy. Nothing here needs a plugin you don’t have, and nothing was uploaded anywhere. I like to build and create things.

I split the file into mid and side, took Welch power spectra (32,768-sample windows, half overlap) and integrated the energy band by band. For the phase question I ran a fourth-order Butterworth low-pass at 120 Hz in both directions, forward then backward, so the filter contributes no phase shift of its own, then took the Pearson correlation of left against right. Across the whole 7:45 that came to +0.993, with 0.5 percent of the sub-120 energy sitting in the sides and a 0.02 dB change when summed to mono.

Then I did it again in five-second windows, because a whole-track average will happily hide a bad drop inside a quiet intro. Ranking the windows by loudness and looking only at the top ten percent, where the bass actually lands, the correlation was 0.999 at both the median and the minimum, with 0.03 percent of the low band in the sides. Loudness came from ffmpeg’s EBU R128 meter: -11.5 LUFS integrated, and a low-versus-high tilt of +10.2 dB, which is the muddy finding above.

What this does not prove: a correlation near +1 says the low end won’t cancel. It says nothing about whether the bass is any good. Those are separate questions and only the first one gets settled by a number. I measured the master rather than the stems, so it tells me the sum behaves, not which element earned it. And a DAW’s correlation meter uses shorter windows than I did, so it’ll twitch where mine reads calm.

Resources

• Jens Blauert, Spatial Hearing: The Psychophysics of Human Sound Localization (MIT Press). The standard reference on how localization works, and why it degrades as frequency drops.

• Attack Magazine, “Mono-Safe Stereo Width”. Why low-frequency stereo content collapses when summed, and how to keep the bottom mono.

• sonible, “Avoiding the Collapse: From Stereo to Mono”. Mono compatibility and where phase cancellation comes from.

Previously in this series: Your Low End Is Lying to You and Who Gets the Downbeat?. The 150 Hz rule I’m revising here came from Width Requires a Strong Center.

Leave a comment

Solo the kick and it hits. Solo the bass and it fills the room. Play them together and the bottom turns to mud. Last time I wrote about why the low end lies to you, how your ears and your room fake the sub. This is the next problem: even when you trust what you hear, the kick and bass can still refuse to agree.

Your hearing becomes less sensitive to deep bass as playback level falls. Your room adds its own fiction by reinforcing some low frequencies and cancelling others. The lesson wasn’t to distrust your ears. It was to stop asking them to work without witnesses. Change the level. Move around the room. Check mono. Use a meter. Listen somewhere else.

But even after you do all of that, you can still build a low end that refuses to hold together.

Maybe the bottom turns soft. Maybe it swells on one note and collapses on the next. Maybe both sounds stay technically audible, but neither feels as solid as it did alone.

Nothing has to be clipping. Neither channel has to be badly recorded. The problem may not be either sound by itself.

It may be the agreement between them.

Turning it down isn’t an agreement

The first move is usually volume.

Pull the bass down a decibel. Then another. The meter calms down, but the mix still feels crowded. Now the same conflict is happening more quietly.

The next move is often a plugin. Saturation. An enhancer. A preset with some version of tight low end in the name. That may add harmonics that help the bass survive on smaller speakers, but it doesn’t necessarily solve the collision underneath. Sometimes it makes the conflict easier to hear. Sometimes it adds more material to an area that was already crowded.

Volume matters. Tone matters. Processing matters.

But before reaching for any of them, I think there’s a better question:

What job is each sound supposed to be doing?

Is the kick carrying the deepest physical hit while the bass supplies movement above it? Is the bass the true sub, with a shorter kick sitting higher? Does the kick own the downbeat while the bass blooms after it? Or are both sounds trying to deliver the same weight at the same instant?

The answer is the beginning of the mix.

Subscribe now

What masking actually means

The ear doesn’t analyze sound as an infinitely precise spectrum display. The cochlea behaves more like a bank of overlapping auditory filters. Energy within and near the same filter contributes to what we can detect, and a stronger sound can make a nearby weaker sound harder to hear. This is one form of auditory masking.

The older Bark or critical-band model gives a useful approximation: below about 500 Hz, critical bandwidths are often estimated at roughly 100 Hz. More recent auditory-filter models describe the details differently, and real masking also depends on sound level, spectral shape, duration and the listener. So this doesn’t mean that everything below 500 Hz becomes one giant perceptual lane, or that two sounds inside the same nominal band become impossible to distinguish.

It means separation becomes harder when strong spectral energy arrives close together in frequency and time.

That matters because a kick and a bass are not single frequencies. A kick may contain a low fundamental or resonance, a changing pitch envelope, body in the low mids and a high-frequency attack. A bass note has a fundamental, harmonics, an envelope and sometimes distortion, modulation or stereo information.

When both sounds place a lot of energy in the same low-frequency region at the same moment, the quieter details of one can be masked by the other. You may still hear that both exist. What disappears is definition.

That distinction matters. Masking is not always one sound erasing another. More often, it is the mix losing the cues that let your ear follow each sound separately.

The low-end right of way

I used to think the cleanest low ends came from giving the kick one frequency and the bass another.

That can help, but it’s too simple.

A techno kick is not parked at 60 Hz. It changes over time. The initial attack may be broadband, the pitch may fall rapidly, and the tail may settle into a much narrower resonance. A bass line changes note, which means its fundamental and harmonics move every time the part moves. A static EQ cut based on one frozen analyzer frame may solve one collision and create another three notes later.

The real decision is not who owns one frequency forever.

It’s who has the right of way when the important parts overlap.

Sometimes the kick gets it. The bass ducks or begins a little later, allowing the kick transient and early body to arrive cleanly.

Sometimes the bass gets it. The kick is shortened, tuned differently or chosen for more attack and less tail so it doesn’t occupy the sub region for half the beat.

Sometimes neither sound needs to move much because their envelopes already fit together. A short kick can leave room for a sustained bass. A long kick can become the bass, with the written bass line living mostly in its harmonics and movement.

This is why two objectively good sounds can make a weak pairing. Quality is not compatibility.

Five ways to settle the argument

Start with selection and tuning. Before repairing the relationship with processing, ask whether these two sounds belong together. A kick with a long, resonant tail and a sustained sub bass may require constant management. A shorter kick, a different bass patch or a small change in decay can solve more than a complicated chain ever will.

Then look at arrangement. Move the bass note. Shorten it. Let it begin after the kick rather than directly under it. Put the slide, accent or loudest note in the space between kicks. A surprising number of mix problems are writing decisions that survived long enough to become engineering problems.

Next, check frequency balance. Listen for the parts of each sound that define its job. A small, broad EQ move may be enough to reduce overlap. Avoid turning an analyzer peak into a law. The important frequency is not always the tallest one. It is the area where reducing one sound makes the relationship clearer without making that sound feel hollow.

Then work on time and envelope. Sidechain compression is one option, but the amount and recovery matter more than the fact that you used it. Duck too deeply or release too slowly and the bass seems to inhale after every kick. That may be the groove you want. It may also be the plugin announcing itself.

If only part of the bass spectrum is colliding, dynamic EQ or multiband ducking can reduce that region when the kick arrives while leaving the rest of the bass more stable. It’s not automatically better than full-band compression. It’s simply a more selective tool.

Finally, check phase and waveform interaction, especially when the kick and bass contain sustained or closely related low-frequency components. Their waveforms add when their instantaneous polarities align and subtract when they oppose each other. Because the bass changes note and the kick changes over time, this relationship is often dynamic rather than one permanent phase setting.

Don’t rotate phase until the analyzer looks impressive. Compare the result across several notes and several bars. A setting that makes one kick enormous may weaken the next one.

Share

The experiment

Mute everything except the kick and bass.

Loop a section that contains more than one bass note. One note is not enough to tell you whether the relationship works.

First, bypass every corrective processor. Listen at a moderate level. Then turn it down. Notice whether you can still follow the outline of both parts or whether they collapse into one pulse.

Now remove the bass. Pay attention to the kick’s attack, body and tail. Bring the bass back and ask which part changed. Did the kick lose its front edge? Did its tail disappear? Did the bass lose pitch definition? Did one particular note cause the whole bottom to surge?

Make one move.

Shorten an envelope. Shift a note. Try a different kick. Add a small EQ cut. Duck the bass by one or two decibels. Change the release. Then bypass the move and listen again.

Don’t judge the fix by how much processing happened. Judge it by whether you can hear both intentions more clearly.

Check in mono, because summing can reveal level and phase relationships hidden by stereo width. Check on headphones and another playback system. A phone speaker can’t tell you whether the sub itself is balanced because it may reproduce very little of it. What it can tell you is whether the kick’s attack and the bass’s upper harmonics still communicate the rhythm and pitch when the deepest octave is gone.

That’s a different test, and an important one.

You’ll know the relationship is working when the kick feels stronger without needing to become louder, and the bass feels clearer without becoming smaller.

Not two sounds fighting for territory.

Two sounds agreeing on the motion.

The kick and bass don’t need to be separated at every frequency or every instant. They need a clear enough relationship that the listener can follow the weight, the pitch and the groove without having to untangle them.

In your own tracks, who usually gets the low-end right of way: the kick, the bass, or whichever one arrived in the session first?

Resources

• Bernhard U. Seeber, “Echoes from the Archives of the Munich School of Psychoacoustics”, Forum Acusticum 2023. A concise academic overview of masking, critical bandwidth and the work of Zwicker, Fastl and Terhardt.

• Brian R. Glasberg and Brian C. J. Moore, “Derivation of Auditory Filter Shapes from Notched-Noise Data”, Hearing Research, 1990. A foundational paper behind modern auditory-filter and equivalent-rectangular-bandwidth models.

• Eberhard Zwicker and Ernst Terhardt, “Analytical Expressions for Critical-Band Rate and Critical Bandwidth as a Function of Frequency”, Journal of the Acoustical Society of America, 1980.

• iZotope, “How to Mix Kick and Bass”. Practical examples of arrangement, complementary EQ, sidechain compression and low-end management.

• Sound On Sound, “Mix Tips for Kick & Bass”. Practical discussion of sound choice, tuning, timing, EQ and dynamics.

Previously in this series: Mixing Bass by Ear: Your Low End Is Lying to You.

Leave a comment

The bass on your track has two settings, and neither one is real. Turn your monitors down to almost nothing and listen to the bass. It nearly disappears. Turn them up loud and it comes back huge, maybe too huge. Same track, same speakers, thirty seconds apart, and the low end changed character completely. Here’s the uncomfortable part: the bass didn’t move. Your ears did.

Now take that mix to a car, a club, a cheap phone. Half the time the low end is gone. The other half it’s a boom that swallows the vocal. Nothing in the session changed. You just finally heard the track somewhere your ears and your room weren’t working together to flatter you.

The low end is the one part of a mix you can’t fully trust yourself to hear. That sounds like a problem. It’s the most useful thing you can know about it.

Two liars: your ears and your room

Your hearing is not a flat measurement tool, and it’s least flat down low.

There’s a well-documented shape to human hearing, mapped almost a century ago, that says we’re much less sensitive to bass than to the midrange. Here’s the part that catches people: the effect changes with volume. Quiet, and the low end falls away. Loud, and it rushes forward. So the balance between your kick, your bass and everything above them isn’t a fixed thing you’re calmly evaluating. It’s a moving target that depends on how loud you happened to turn the knob. Mix quiet and you’ll overcook the bass to compensate. Mix loud and you’ll starve it. Most “my low end doesn’t translate” stories start right here, at a monitoring level nobody chose on purpose.

Then there’s the room, which is the louder liar of the two.

Subscribe now

Low frequencies are long. A 40Hz wave runs about 28 feet crest to crest, longer than most rooms people mix in. Waves that size don’t politely arrive at your ears and stop. They bounce off the walls and pile back onto themselves, and where they pile up you get a boom, and where they cancel you get a hole. Those peaks and holes can be huge, often ten to twenty decibels apart, sometimes more. Move your chair a foot and the bass note you were just trusting can double or vanish. Your listening position is parked in one spot in that mess, telling you a story about the low end that’s true nowhere else, including the club you made it for.

Under both sits a simpler fact. The lowest bass is felt as much as heard. On a real system it lands in your chest and the floor, not just your ears, and a small room at mixing volume gives you almost none of that. You’re judging with a sense you can barely use, in a space that warps it, at a volume that changes the answer.

I’ve started calling that bottom stretch the blind octave. It’s the region, very roughly the sub and low bass, where you aren’t really hearing so much as guessing with better equipment.

The false assumption is that a weak or muddy low end is a sound problem: wrong plugin, wrong sample, not enough sub. Sometimes it is. Far more often the sound is fine and your read on it is broken.

Stop trusting the octave you can’t hear

Here’s the shift. You don’t fix the blind octave by listening harder. You fix it by refusing to decide it on your ears alone.

Good low-end mixers are a little paranoid down there. They treat the sub the way a pilot treats a cloud: don’t trust your body, trust the instruments, because your body is the exact thing being fooled. That means bringing in witnesses that don’t care how the room feels tonight.

Mono is the first witness. Collapse the mix to mono and a lot of low-end trickery falls apart, because most big club systems and every record cut to vinyl sum the bass to mono anyway. If your sub carries stereo width and part of it is out of phase, it can cancel toward silence on exactly those systems. If it dies in mono, it was never really there.

A meter is the second witness. Not as the truth, but as a second opinion that the room and the hour can’t influence. When your ears say “huge” and a spectrum analyzer says the 50Hz region is quietly eating ten decibels of headroom, trust the argument between them, not either one alone.

The phone is the third, and it teaches the strangest lesson. A phone speaker can’t move real sub at all, and yet a well-made bass still reads as low on it. That’s because your brain rebuilds a missing low note from its harmonics, the overtones sitting up in the range a tiny speaker can reproduce. So part of translating the low end is making sure something lives an octave or two up that implies the weight, even when the weight itself can’t get through. The sub is for the club. The harmonics are for everywhere else.

None of these is the answer alone. Together they fix a position your ears can’t give you by themselves. The sounds you can genuinely hear and verify are the sounds you can control. The rest is luck wearing a big subwoofer.

Share

The experiment: catch your ears in the act

Do this before your next low-end decision. It takes ten minutes and it’s a little humbling.

First, the volume test. Loop your busiest low-end bar. Set what feels like a normal level, then turn it down until it’s almost too quiet to enjoy, then up until it’s slightly too loud. Watch the kick and bass swell and shrink against the rest of the mix as you move the knob. Find the moderate level where the balance stops lurching around, somewhere near a normal conversational loudness, and make your low-end calls there. Not quiet, not blasting. The honest middle.

Second, the room test. Hold or loop one sustained sub note and slowly move your head around your seat, then stand, then crouch. Listen for the note ballooning and dropping out as you go. You’re not fixing anything yet. You’re proving to yourself that your chair is one lie among many, so you quit over-trusting it.

Third, the witnesses. Sum to mono: does the low end survive? Pull it up on your phone: does the bass still imply itself through the harmonics? Write one sentence per check. Do that for a week and you’ll stop arguing with your monitors and start reading them.

You don’t win the low end by hearing it better. You win it by knowing exactly when your ears are lying, and having somewhere else to look.

What volume are you actually at when you commit your kick and bass, and have you ever checked whether that’s the level telling you the truth?

Leave a comment

The session is open. Every plugin you own is one click away, every sound you could possibly make is on the table, and you’ve been sitting there for twenty minutes without committing to anything. You audition a pad, then three more, then a different pad in a different synth. Nothing’s wrong with any of them. That’s the problem. When everything is possible, nothing is necessary, and the track just sits there waiting for you to decide who it is.

Now think about the last track that came out fast and sounded the most like you. You probably weren’t choosing from everything. You were working inside a narrow set of moves you trust, almost without noticing you were doing it.

That wasn’t luck. That was a system, and you already have one. You likely just haven’t written it down.

The false freedom

The story we tell ourselves is that more options mean more creativity. More sounds, more plugins, a wider palette, and the good ideas will have more room to happen.

Some of that is true. Sometimes you genuinely need a sound you don’t own yet, a new technique, a wider range. I’m not going to pretend limitation is always a virtue. The room can matter. The tools can matter.

But past a point, more choice stops feeling like freedom and starts feeling like a blank page with too many doors. Every one of them open. None of them chosen. You spend your energy deciding what to do instead of how to say it, and the track pays for it.

Stravinsky said it better than I can: the more constraints you impose, the more you free yourself. Brian Eno and the painter Peter Schmidt built a whole deck of cards around the same idea back in 1975, little instructions you draw at random just to close a few doors on purpose. The research backs the shape of it too, with one honest caveat: it’s the right few constraints that help. Pile on too many and you choke the work just as badly as leaving every door open. The skill is choosing which ones.

What a system actually is

A system isn’t a preset and it isn’t a genre. A preset is somebody else’s decision frozen in place, and a genre is a crowd. A system is the small set of things you keep doing on purpose, and the smaller set of things you refuse to do at all.

You’ve spent this whole season building the parts without calling them that.

The elements you can’t stop reaching for, the ones that showed up in track after track before you ever decided they were yours: that’s your signature. The frequency range you live in, and the range you leave to someone else: that’s your territory. The one or two sounds a track can’t survive without: that’s your necessity. And the differences you’ve trained yourself to actually hear, the gaps you can name now instead of only feel: that’s the ear doing quality control on all of it.

Four discoveries. On their own they’re insights. Written down together, on one page, they become a system.

I think of it as designed scarcity. Not scarcity because you’re broke or your rig is small, but scarcity you chose, on a calm day, because you know it makes you move.

There’s a twist in this. The constraint is the thing that makes you recognizable. Nobody remembers the producer who can do everything. They remember the one whose tracks share a spine you could pick out with your eyes closed: the low end that always sits a certain way, the space that always feels the same size. That spine is just a constraint, held long enough that it hardened into a signature.

Share

Build the page

Here’s the drill. It takes an afternoon and it changes how the next fifty tracks feel.

Open a blank note. Give it two lists.

The keep-list is the moves you reach for in almost every track. Your kick and bass relationship. The saturation curve you always land on. The reverb you trust. The three synths you actually finish things in. Be honest, not aspirational. Write what you do, not what you think a serious producer is supposed to do.

The refuse-list is the doors you’re closing on purpose. The frequency range you won’t crowd. The plugin you keep buying and never improving on. The tenth layer you always add and always regret. The genre move that isn’t yours, no matter how well it’s working for everyone else this year.

Then produce one track using only the page. No shopping. When you feel the urge to audition a fourth pad, you don’t. You reach for the one on the list and you make it work, which is where most of the actual craft has been hiding the whole time.

Notice two things afterward. How much faster you moved. And how much more it sounded like you, without you trying to make it sound like anything.

The thing you can’t generate

There’s a reason this matters more now than it did five years ago.

Anyone can generate a ten-layer, full-spectrum, technically clean track from a sentence. The floor came up. Competence got cheap. What didn’t get cheap is a point of view, because a generator doesn’t have one. It has an average. It will hand you everything, which is exactly the problem you started the session with: everything, and no reason to choose.

Your system is the reason to choose. It’s the running record of what you keep and what you refuse, and that isn’t something you can prompt for, because it isn’t information. It’s taste with enough reps behind it to harden into a shape.

That shape is the most valuable thing you own as a producer. Not because it’s human and therefore automatically better. Because it’s yours, and you picked it on purpose.

You don’t earn a sound by keeping every door open. You earn it by knowing exactly which few you walk through, and closing the rest without flinching.

What’s the move you reach for in almost every track, the one you’ve never actually written down?

Leave a comment

You bounce the mix, drop it next to a track you love, and switch back and forth. Yours sounds smaller. Flatter. A little amateur in a way you can’t put your finger on. You know something is wrong. You just can’t say what.

So you start guessing. Nudge the highs. Add a touch of reverb. Pull the vocal up half a dB, push it back down. An hour later the mix is different and no better. You’re tired, your ears are cooked, and now everything sounds suspicious.

So you blame the usual suspects.

The room. The monitors. The plugins. The headphones. The converter. The last YouTube tutorial that convinced you your whole gain staging philosophy was broken.

And yes, sometimes it is the room. Sometimes the monitors are lying to you. Sometimes a plugin is the wrong tool, or the right tool used in the wrong way. All of that matters.

But this piece is about the problem underneath those problems.

Because even in a treated room, on good monitors, with the right tools open, you still have to hear the difference clearly enough to name it. If all you can say is “mine sounds worse,” you’re not mixing yet. You’re guessing with better equipment.

The gap you can feel and can’t name is one of the most important measurements in your whole studio, and it isn’t only coming from your gear.

It’s coming from your perception - your ability to hear.

What Nobody Tells You About Hearing

The false comfort is that mixing is a knowledge problem. Learn the frequencies, memorize what a compressor does, buy the monitors the pros use, and clarity will follow.

It likely won’t. And you probably already have the proof sitting on your drive.

You own reference tracks that sound incredible. You can hear that they sound better. You know the vocal sits in the track differently. You know the low end feels more controlled. You know the whole thing has more depth, width, punch, air, whatever word your brain reaches for first.

But knowing that it sounds better is not the same as hearing why it sounds better.

That’s the gap.

The problem is not that you don’t care enough, or that you haven’t watched enough tutorials. The problem is that your ear may not yet be trained to separate one kind of difference from another. Low-mid buildup, harshness, masking, transient shape, stereo width, reverb length, vocal level, arrangement density - they can all collapse into the same vague feeling: “mine sounds worse.”

That vague feeling is real. It’s just not clearly useful yet.

This is where perception comes in. Not perception as some mystical producer gift. Perception as the ability to hear the mix clearly enough to identify what is actually happening inside it. To move from “something feels off” to “the low mids are crowding the kick,” or “the vocal is bright but not present,” or “the reference feels wider because the sides are carrying different information.”

That kind of listening belongs mostly in what I call the diagnostic stage. Not the composition phase, when you’re trying to catch the first idea or follow the melody that just showed up. Not the arrangement phase, when the track is still telling you what it wants to become.

Let the writing phase be alive.

But when you start comparing, revising and mixing, you need a different mode. You need to stop reacting to the whole emotional blur and start hearing the parts inside it.

The information was never the real bottleneck.

The bottleneck is whether your ear can hear what is happening inside the mix clearly enough to turn that vague feeling into a diagnosis.

Why This Is Good News

Here is the best news in mixing: perception is not a gift you were or weren’t born with.

It’s trainable.

Not in the vague motivational sense. In the dull, literal, repeatable sense. Focused attention on one specific feature changes what your brain can pull out of the blur.

That’s perceptual learning. You are not just learning a fact about sound. You are training recognition. You are building a category your ear can use later.

That is why a radiologist can see something in a scan that looks like grey fog to you. It is why a sommelier can taste oak, acid and structure where someone else just tastes “wine.” They did not grow better eyes or a better tongue. They trained themselves to notice distinctions that were already there.

Mixing works the same way.

At first, low-mid buildup is just “mud.” Harshness is just “too bright.” Masking is just “crowded.” Reverb length is just “washed out.” The ear feels the problem, but the brain has not separated it into a useful category yet.

Once the category exists, the same sound gives you more information.

That is the strange part. The speakers did not change. The room did not change. The plugin did not suddenly become smarter.

You changed what you were able to hear.

Producers think they’re chasing better gear. Sometimes they are, and gear does matter. But a lot of the time, they’re waiting to perceive what is already coming through the speakers or headphones.

Hearing Is Naming

There’s a loop here worth slowing down on.

You can feel what you can’t name. You just can’t work with it very well yet.

That is the problem with the vague “mine sounds worse” feeling. It may be pointing at something real, but until you can separate it, name it and test it, it stays slippery. Low end? Masking? Too much reverb? Not enough transient? Wrong arrangement density? They all feel bad when the ear has not learned to tell them apart.

So the work is not just listening more.

You’ve already heard thousands of hours of music. Your mixes did not fix themselves. Passive exposure can build taste and familiarity, maybe. It does not reliably build diagnosis.

The work is listening to one thing on purpose, putting a word on it and checking whether the word was right.

That’s the engine: attention, language, proof. In psychology, this sits close to perceptual learning and category formation. In the studio, it feels much less clinical. One day the blur has a name, and once it has a name, it starts showing up everywhere.

Do that enough times on the same feature and it stops being effortful. The 250Hz mud starts announcing itself before you reach for an EQ. The too-long reverb tail feels obvious. The vocal that seemed “fine” suddenly sounds buried by half a dB, or bright in the wrong place, or disconnected from the room around it.

You’re not memorizing facts.

You’re growing sensors.

The first one is always the strangest. Say it’s low-mid boxiness, that cardboard honk somewhere around 300 to 500Hz. For months, it’s invisible to you. Then someone makes you solo it, sweep it, sit with it, name it and compare it against the reference. A switch flips.

Now you hear it everywhere.

In your mixes. In your friend’s mixes. In songs you’ve loved for twenty years and never once noticed it in.

That’s the tell that a sensor is real. You can’t turn it back off.

Share

The Week-Long Drill

Pick one reference track you wish yours sounded like. Just one. Not ten. Ten references will turn into taste soup.

Bounce your current mix. Level-match it against the reference as honestly as you can, because louder will almost always trick you into thinking “better.” Then, for the next week, ten minutes a day, run the same small ritual.

A/B them.

The instant you feel the gap, stop.

Don’t fix anything yet. Name it first, out loud or write it down if you can. Use the most specific sensory language you can force out.

Not “mine sounds worse.”

Say “mine feels boxy in the low mids,” or “the reference has air up top that mine doesn’t,” or “their vocal sits forward and mine feels buried.”

The exact word does not have to be perfect at first. Maybe you call it boxy. Someone else calls it cardboard. Someone else says the vocal feels trapped in the speaker. That’s fine. Early on, the word is just a handle.

But over time, the handle has to attach to something real.

That is where shared vocabulary helps. If you work with other producers, engineers or vocalists, words like muddy, harsh, boomy, nasal, airy, forward and buried become a common map. Not because everyone uses them with laboratory precision, but because they give the room a place to start.

Your private language helps you notice. Shared language helps you collaborate. The goal is not to necessarily sound like an engineer. The goal is to connect the word to the sound clearly enough that you can test it.

If you can’t get specific, guess.

A wrong specific guess teaches you more than a true vague one, because tomorrow you can test it.

Then isolate the thing you suspect. Solo the element if that helps. Sweep a narrow EQ band until the thing you named jumps forward. Cut it and see if the feeling changes. Shorten the reverb and listen for whether the vocal moves closer. Pull the kick down half a dB and notice whether the bass stops folding under it.

Then go back to the full mix.

That part matters. Solo can help you find the feature, but the full mix is where the truth lives. A sound can be ugly alone and perfect in context. It can also sound fine alone and fail the second everything else comes back in.

You’re not trying to finish the mix during this drill. You’re closing the loop between the word and the sound.

Three named differences a day. Same reference all week.

By the end of the week, something usually starts to shift. You open a new session, hit play, and the boxiness you’ve been hunting is just there. Not mystical. Not dramatic. Just more obvious than it used to be. Named before your hand reaches for anything.

That’s a new sensor coming online.

You don’t get to un-hear it now.

What No One Can Hear for You

Worth saying plainly in this particular year: a teacher, framework, analyzer or model can suggest the move now.

Cut 250Hz. Shorten the reverb tail. Tame the harshness around 3k. Check the masking between the kick and bass.

Sometimes that suggestion will be useful. Sometimes it may even be right.

But the suggestion is not the skill.

The skill is hearing whether the move actually helped. Did the vocal get clearer, or just thinner? Did the low end tighten up, or did the track lose weight? Did the reverb step back into the mix, or did the whole thing suddenly feel smaller?

That judgment still has to happen in your body, through your ears, against your taste and the track you’re actually trying to make.

Other people, and even AI, can point at a possible problem. A mentor can give you language. A spectrum analyzer can show you a clue. A framework can give you a place to start.

That’s not nothing.

But none of it can become the part of you that knows what “better” sounds like.

Taste is perception with enough reps behind it that it starts to feel like instinct. You can borrow a suggestion. You can’t outsource the listening.

The ear is the one instrument in the whole chain that has to be yours.

The sounds you can hear are the sounds you can control.

Everything else is luck.

So here’s the real question:

What’s the problem in your mixes you can feel every single time and still can’t put a word on?

Leave a comment

The track was working. Eight elements, all sitting right, and you added a ninth because you had the channel open. Now nothing lands. The kick feels smaller. The lead you loved an hour ago sounds ordinary. Nothing in the mix actually broke. You just filled the last open seat in the room, and now everyone’s talking over each other.

Here’s the part that stings. A machine will generate a finished, full-spectrum, ten-layer track in seconds now. Competent. Loud. And forgettable, because nothing in it was chosen and nothing in it would be missed. The one move the machine doesn’t make is leaving something out on purpose.

That move is the whole game.

The More-Is-Richer Reflex

Most of us learned production by addition. A track felt thin, so we layered. Another pad, a second percussion loop, a counter-melody, a noise riser to cover the transition. Each layer solved a small problem in the moment. The reflex is honest: when something isn’t working, do more.

The trouble is that richness and density aren’t the same thing. Density is how much is playing. Richness is how much each thing means. You can raise density forever. Richness has a ceiling, and you hit it the second the listener can’t tell what to follow.

A full arrangement isn’t a rich one. It’s a loud room.

What the Ninth Sound Actually Costs

Attention is the budget. Track count isn’t.

There’s a working rule among arrangers that the ear can really hold about three things at once: a foreground, a counter-voice and a bed. Push past that and the brain doesn’t hear “more.” It starts triaging. It drops something so it can follow the rest, and you don’t get to pick what it drops.

So the ninth element isn’t a ninth layer of interest. It’s spending attention that used to belong to your best sound. Every layer you add is a quiet tax on the one thing you most wanted heard. The lead got smaller and you never touched its fader.

Think of a photograph with one subject and a clean background. The eye goes straight home. Now fill every inch of that frame with detail, all of it technically sharp, and the subject has to fight the wallpaper to be seen. Nothing got brighter. The center just lost its claim.

(I spent years adding. I had a habit of dropping a second, quieter percussion loop under the main groove on nearly every track, just for “movement.” One night I muted it on a tune that wasn’t sitting right, meant to put it back later, and forgot. The track got stronger. The groove I’d been propping up had been load-bearing the whole time, holding up nothing but my own nervousness about empty space.)

The Removal Test

You don’t find the essential sound by listening for what’s good. You find it by taking things away until something dies.

Open a track you love. Mute one element. Play eight bars. Then ask the right question, not the easy one. Not “is the track better without it.” Ask “could this be gone and the song still be itself.” If yes, that sound was furniture. Useful, maybe. Not essential. Leave it muted and move to the next one.

Keep going. Most elements will pass quietly: gone, and the track shrugs. Then you’ll pull one and the floor drops out. The song stops being that song. That’s the sound you can’t remove. That’s the one carrying the identity.

Now run it on your own last three tracks.

You’re listening for two things. First, how many of your elements are furniture you’ve been mixing around for years. Second, whether the irreplaceable sound is the same kind of thing each time: a texture, a voice, the one move that keeps showing up as load-bearing. That repeat is the fingerprint from two weeks ago, seen from the other side. Last time you found it by what you keep adding. This time you find it by what you can’t take out.

Same signature. Read from the negative.

And once you know it, you can stop discovering it by accident. The usual order is backward: we stack eight elements and hope the identity turns up somewhere in the pile. Try it the other way. Start the next track with the one sound you already know you can’t remove, and add only what that sound asks for. Most of the nine never get invited. You finish faster, and the thing you finish already sounds like you.

Share

What the Machine Can’t Subtract

Generation is addition by nature. Ask for more and it gives you more, instantly, in full spectrum, with no fear of the empty bar. It will never leave a hole, because it has nothing to lose by filling one.

Subtraction is a decision, and a decision needs someone with taste and something at stake. Removing the ninth element means betting that the eight that remain are enough, and standing behind the bet. You can’t prompt for that. It’s the part of the work that still asks you to have an opinion about what matters.

So when everything around you can be infinitely full for free, full stops being impressive. What reads as a living hand is restraint: the sound that’s clearly there because you chose it, and the silence around it that’s clearly there because you guarded it.

The track that sounds like you isn’t the one with the most in it. It’s the one you refused to crowd.

Build tracks with only the sounds you can’t remove.

When you mute your favorite element, does the song just get quieter, or does it stop being yours?

Leave a comment

The plugin demo promised full spectrum response. Your mastering chain is catching everything that gets too loud. You’ve A/B’d your reference tracks and confirmed there’s energy across the whole frequency range.

You’ve filled the spectrum.

You’ve also made yourself interchangeable with everyone who did the same thing.

The producers with the most immediately recognizable sounds aren’t the ones who found the right plugin to cover every frequency band. They’re the ones who decided which frequencies are theirs and let everything else stay quiet or at least settle into the background. The absence isn’t a gap in the mix. It’s part of the strategy.

The Full-Spectrum Myth

Most mixing guides treat a balanced, full spectrum as the technical goal. It makes sense as a broadcast standard: you want energy where it should be and nothing extraneous. Gaps feel like problems. Thinness reads as incomplete.

That framing is correct for broadcast. It’s the wrong target for identity.

The producers who sound most like themselves aren’t the ones whose spectrum analyzer reads flat across the full 20Hz-20kHz range. They’re the ones whose mix has a specific spectral posture: deliberate presence in certain ranges, conspicuous absence in others. That posture is as identifying as a vocal fingerprint. When you hear the first two bars of a Jon Hopkins or Underworld production, you’re not hearing a full spectrum. You’re hearing a claimed territory.

You know whose it is before the drop.

The frequency bands they don’t occupy aren’t oversights. They’re what makes the occupied territory feel like it belongs to someone. A mix that’s trying to be everywhere is a mix that belongs to no one.

What the Empty Bands Are Doing

Empty frequency ranges aren’t neutral. They’re working.

Think about negative space in a photograph. A portrait against a clean background isn’t missing information: the empty space around the subject focuses attention, adds contrast, makes the face matter. Fill every corner of the frame and the subject competes with the background. The image loses its center.

Frequency space works the same way.

When a producer consistently leaves the upper 10kHz region sparse, that’s not a bandwidth limitation. The sparse high end is changing everything underneath it: the mids read warmer, the low end feels more dominant, the overall texture sits in a room differently than a mix that’s bright all the way up. That spectral posture communicates something. Listeners can’t describe it. They recognize it.

What fills the spectrum is what you’re showing. What stays empty is what you’re refusing to compete for. Both decisions define the territory.

(I spent two years trying to make my high end brighter because reference tracks in my genre were brighter. I kept reaching for air shelf EQ, brighter cymbals, more presence in the 8-12kHz range. My mixes got cleaner. They stopped sounding like mine. Eventually I backed off, and within a few sessions the identity came back. The difference was in what I stopped adding, not what I found.)

The Notch Test

Load a spectrum analyzer. Pull up a track by a producer whose sound you find immediately identifiable: someone where you’d know it in the first 10 seconds even without seeing the name.

Let the track play and watch the areas that aren’t moving. Find the frequency bands where almost nothing lives. Not the transient peaks that come and go. The consistent quiet zones. The places they don’t go.

Those aren’t problems in the mix. That’s their spectral territory, defined in part by what they chose not to claim.

Now apply systematic notch filtering: pull down 5kHz of bandwidth at a time and listen. Don’t listen to the cut itself. Listen to what the cut reveals about what was there. When you pull a band the producer has actually claimed and something irreplaceable leaves (that warmth, that edge, that specific texture), you’ve found it.

Do this for three tracks by the same artist. The same bands will hold the character across different productions.

That overlap is their identity, written in frequency.

Now open one of your own productions and run the same test. Where do your mixes consistently live? Where do they consistently leave space? The pattern in that answer is the spectral version of the fingerprint from last week: already there, already accumulated, whether you named it or not.

Claiming Your Range

The goal isn’t to strip three frequency bands and call it a sound. It’s to understand what you actually occupy when you’re not overthinking it, then make that choice with intention.

Most producers have a spectral tendency they’ve been building for years without naming it. The 800Hz warmth that keeps appearing. The way their high mids sit at 3-4kHz, present but not sharp. The sub that occupies a specific part of the low end rather than fighting across the whole 20-80Hz range.

That tendency is information. Once you see it, the question changes: not “why does my mix sound like this?” but “which frequencies do I want to own so completely that pulling them would pull me out of the track?”

That’s a different way to make decisions in a session. Not balancing everything. Asking: what is mine, what am I keeping, and what belongs to silence?

The spectrum you don’t use is half of your identity. Start treating it like a choice.

Share

Own your frequency range. Let silence own the rest.

Which frequency band in your mix is so consistently occupied that removing it would remove something that sounds like you?

If you want to hear this in practice, ASCENDANCE (Original Mix) is the clearest example from my recent work: the decisions about what occupies the 2-4kHz range and what I left alone in the upper frequencies were made with this exact frame in mind.

Spotify · SoundCloud · YouTube

Leave a comment

Pull any three of your sessions from the last year. Open the effects chains. Look at the reverb on the pads, the saturation on the synths, the space around the main melodic element.

Do that for three tracks. Then five. Then ten.

At some point you’ll see it: the same choices appearing across different productions. Not exactly, but close enough to be undeniable. The same reverb tail character. The same saturation texture on the mid-range. The same way you tend to treat the high-frequency content of your leads. Perhaps you didn’t plan that. But there it is.

That’s your fingerprint. And you’ve been building it for years without knowing.

What Most Producers Believe About Signature

Most producers think a signature sound is something you design.

You research the artists you admire. You figure out which processing chain creates that specific texture. You work backward from the sound you love to the method behind it. You experiment deliberately until you find something that feels distinctly yours. Then you build on it.

That framing is backward.

The producers I know who have the most immediately recognizable sounds didn’t get there through deliberate construction. They got there by making the same choices repeatedly: the same harmonic preferences, the same spatial decisions, the same tolerances for density and space, until those choices accumulated into something coherent.

Your signature wasn’t designed. It was deposited. Session by session, choice by choice, the same way a path gets worn into a field: not because anyone planned the route, but because everyone kept walking the same way.

What’s Actually Happening

Here’s the reframe: your sonic fingerprint isn’t something you build toward. It’s something you’ve already built without noticing.

Every session contains an archive of decisions. The reverb you reached for when the track needed air. The saturation you applied when something felt too thin. The way you tend to handle the sub range when the kick and bass are competing for the same space. These decisions happen fast, often intuitively, and you make the same ones repeatedly because they express something real about how you hear sound.

Over 20 or 30 or 50 sessions, those decisions accumulate. The pattern emerges before you recognize it.

This matters because most producers spend time trying to invent something that’s already there. They chase trends to sound more current, imitate influences to sound more polished, and in the process sand down the very thing that makes them distinct. Your fingerprint is in the choices you make when you’re not thinking about what you should sound like.

Why Recognition Changes Everything

There’s a practical difference between making unconscious choices and making the same choices consciously.

When you don’t know what your fingerprint is, you can’t strengthen it. You can’t ask “is this decision consistent with what I actually do?” You can’t know when you’re drifting toward someone else’s sound versus deepening your own.

Once you see it, you can make it deliberate. The reverb that kept appearing in your sessions: choose it with intention. Not because it’s familiar, but because it’s yours. The harmonic texture you kept gravitating toward: now it’s a tool, not an accident.

This is the difference between a producer who sounds like themselves on good days and one who sounds like themselves consistently. The second one knows what they keep returning to and returns to it on purpose.

The Overlap Test

Block out 30 minutes. Do this in order.

Pull your last 10 tracks. Finished or not, it doesn’t matter.

For each one, open the effects chain on three elements: a pad or texture sound, a melodic lead and the main rhythmic element. Don’t overthink which elements. Just the first three your eye lands on.

For each element, write down four things: the reverb type and approximate tail length (long and lush, short and tight, somewhere in between); the saturation character (harmonic, tape, clip; describe what it sounds like, not the plugin name); the high-frequency treatment (present, rolled off, bright but narrow); and the low-mid density (thick, hollow, tight, pushed).

Do that for all 10 tracks.

Then look at the columns. Not for exact matches. For tendencies. Where does the reverb tail land most of the time? How does the saturation character tend to sit? What do you do with low-mids when you’re not actively thinking about them?

The pattern in those tendencies is your signature. Not as a brand concept. As an actual sonic reality you can work with.

(I ran this exercise a few years ago and found I consistently over-treat the 2-4kHz range with a specific saturation character on almost every production. I’d thought it was a track-by-track decision. It wasn’t. It was the decision I always made. Once I saw it, I could choose when to apply it and when to leave it alone. The result was more controlled than anything I’d done before.)

Share

What You Do With It

The point of finding your fingerprint isn’t to mechanically repeat it.

It’s to stop accidentally erasing it.

Most producers have a phase where they work hard to sound more like their influences. That’s useful early: you’re building vocabulary, understanding what’s possible, training your ear for what you want. But at some point the work shifts. You’re not learning language anymore. You’re finding what you have to say.

The overlap test won’t tell you what to do next. It’ll show you what you already do. What you do with that is where the interesting decisions start.

Your fingerprint got there because something in you kept making those choices. Not by accident. By instinct. The work now is understanding that instinct well enough to use it on purpose.

You didn’t invent your style. You discovered what you keep returning to.

What’s the one sound design element you’ve found appearing in almost everything you make, even though you never consciously planned it that way?

ASCENDANCE (Original Mix) is the most recent track where I made these choices deliberately: the saturation character, the reverb treatment, the way the mid-range breathes. If you want to hear what a discovered fingerprint sounds like when it’s been noticed and then used on purpose, start here.

Listen on Spotify

Listen on SoundCloud

Experience on YouTube

Leave a comment

You’re at minute 90 of a four-hour set. The floor has been moving for an hour. The DJ is still doing everything right: tempo locked, transitions clean, energy sustained.

The room isn’t responding the same way it was at minute 30.

Not inattentive in the phone-checking way. Something subtler. The kick lands and the floor absorbs it. Nothing comes back. That’s not fatigue from effort. That’s the brain running a pattern it already knows. When it knows what’s coming, it stops paying forward.

The Energy Myth

Most producers finishing a track for DJ use assume the same thing: if it’s energetic enough, it will land. Build the drop harder. Push the sub. Drive the bassline. Energy is the currency of a DJ set.

Energy maintains a floor. Novelty moves it.

Neural engagement with repeated musical patterns drops as the brain learns to anticipate what comes next. The effect isn’t gradual. It’s a cliff. At some point, more intensity doesn’t produce more response. It produces less, because the brain has already modeled what comes next and is no longer firing in recognition of the unexpected.

You’ve felt this in a room without knowing what you were feeling. The floor peaks. It plateaus. More tracks come in with the same rhythm profile, the same harmonic texture, the same drop weight. The response stops scaling with the intensity. The DJ isn’t losing the room. The production window closed.

That ceiling isn’t volume. It’s anticipation.

What Novelty Actually Is

Novelty in a set isn’t a genre switch or a BPM jump. It’s a production character change.

Think of it like a film that uses the same camera angle for 40 minutes. Nothing is technically wrong with any individual shot. The eye adjusts, then goes numb. Then a cut changes the angle and the scene suddenly has depth again, not because the story changed, but because the visual grammar shifted.

A set works the same way. What shifts the grammar isn’t energy level. It’s the production character of what enters the mix.

The pad texture that hasn’t appeared in this set before. A kick with a different transient shape than the last five. A harmonic layer occupying a frequency range nothing has touched in 20 minutes. A decay length that doesn’t share its profile with everything surrounding it.

The listener’s brain can’t articulate what changed. It feels the shift. Attention reinstates.

That reinstatement is the floor “waking up” between tracks: the moment when something lands differently than expected and the room gives something back before it should. The DJ’s crossfader didn’t do it. Your oscillator choices from three months ago did.

The Window

Listen to a recorded four-hour set with this specific attention: not the transitions, not the track selection, not the BPM arc. Listen for when the production character shifts. New textural layer. Different harmonic posture. A rhythm profile that changes before the drop arrives. A sub relationship that doesn’t match the previous three kicks.

Mark each shift. Count the tracks between them.

The window runs somewhere between five and eight tracks before something in the production vocabulary of the set changes, or the floor stops giving back what it’s being asked to feel. That’s not a law. It’s an observation from repeated listening to sets that hold floors and sets that lose them. But it’s consistent enough that experienced DJs and producers feel it without being able to name it.

When the character shift arrives, the room wakes up. Not because it got louder. Because the sonic vocabulary became unpredictable again.

Your Production as Set Inventory

Every finished production is one unit of set inventory. It contributes a specific texture, harmonic posture and rhythm profile to whatever window it occupies in a set.

The problem: you can’t control the window.

A DJ might play your track after four cuts with a nearly identical pad texture and a similar kick weight. In isolation, your production is distinctive. In that window, it’s continuation rather than character shift. The floor’s response tells you which context it’s in. You weren’t there for the decision.

What you can do: produce for distinctiveness that survives context.

The elements that hold their ground in a window of similar productions sit in distinctly different places. A lead synth claiming a frequency range nothing else in your genre commonly uses. A rhythmic subdivision that doesn’t share its pattern with the standard 16th-note loop. A textural layer in the sub-mid (around 150-250 Hz) that creates a body resonance the surrounding tracks don’t have. A decay that’s half the length of what preceded it, or twice as long, creating time contrast where the ear expects continuation.

These aren’t accidental. They’re production decisions made in the session when you stop asking “does this sound good?” and start asking “does this sound different from what’s most likely to surround it?”

Share

The Listening Experiment

Pull a recorded four-hour set you didn’t produce. Three to four hours is better than one.

Don’t listen for mixing technique. Listen for production character shifts: the moments where the sonic vocabulary changes. Mark each one.

Now count the tracks between marks.

You’ll see the window. Sometimes five songs. Sometimes eight. Rarely more than ten before the next character shift arrives. The DJs who hold floors across four hours aren’t only technically skilled. They’re moving through a novelty schedule, consciously or not, because the best track selection instinctively finds it.

Now pull your last finished production. Mentally surround it with five tracks most likely to precede it in a set from your genre. Does your production read as a character shift or as continuation?

That’s the right question. Not “is this good?” The floor at minute 90 doesn’t answer that question the same way your monitors do.

Capstone

Recently we started with a single reframe: a set isn’t a long song. It’s a different compositional form with different production rules. A track in a set isn’t a standalone statement; it’s a movable element in a larger narrative.

From there, every subsequent article addressed a different dimension of that narrative. The key you choose determines what harmonic conversations are possible downstream. The breakdown density you build determines how much tension can accumulate before the drop. The mix you deliver will be rewritten by every room it plays in, and anticipating that rewrite is part of the production job.

And now: the production you contribute is one unit of set inventory, and its job isn’t just to sound good in isolation. Its job is to shift the production character of the window it enters.

You’re not making standalone tracks. You’re building elements for a narrative arc you don’t fully control. The session is where you decide how much weight your contribution carries when the context is someone else’s.

Your production is set inventory. Account for what’s already playing.

What’s the one sound in your current production that no other track in your genre regularly has? If you can’t name it, the floor at minute 90 won’t feel it either.

Leave a comment

The kick sits perfectly in the bedroom. Sub full without being muddy. Transients with the kind of snap that makes you lean into the monitors a little. You’ve A/B’d it against reference tracks and every check tells you it’s right.

Then a DJ plays it in a club.

The kick is gone. Not quiet. Gone. The sub is an indistinct wall of pressure. Something in the 2-4 kHz range, where you carefully carved space, is cutting at a volume that makes people near the speakers take a half-step back.

Same production. Same mix. Different room.

The Neutral Room Myth

The assumption most producers carry: the room is passive. You finish the mix; the room just plays it. Translation rituals, mono checks, car tests: those exist to verify the mix is portable. If it survives a phone speaker, it’s fine.

It isn’t.

Your production doesn’t just play in a room. The room processes it. Every listening environment has its own resonant frequencies and reflection patterns. The room reinforces some frequencies and kills others. How reflective the surfaces are determines whether decays stretch out or stop short.

A club doesn’t play your mix. It rewrites it in real time, differently from your bedroom, the car, the DJ’s monitoring rig, and every other venue the track will travel to.

What the Room Is Doing

Your kick sounds punchy at home, in part, because your room is quietly reinforcing the frequency range it sits in. Standing waves build at specific frequencies based on room dimensions. Your bedroom has its own set. A club has a completely different one. The low-end boost your room was providing (the thing making your kick feel full and present) doesn’t come with the track. It stays in your room.

The midrange is less predictable. Hard parallel walls create comb filtering: peaks and notches across 1-4 kHz where reflections arrive at the listening position slightly out of phase. A vocal that sounds intimate and forward in headphones can turn harsh and fatiguing in a mid-size club. Not because the vocal changed. Because the room’s reflection pattern is smearing the frequencies that made it feel close.

Highs are the most room-dependent of all. A venue lined with acoustic panels swallows the top end that makes a pad shimmer. A concrete room scatters it into a bright, fatiguing smear. The shimmer that made your pad feel open in headphones can turn into a 10 kHz edge that makes people wince by the second chorus.

None of these are mix errors. They’re room errors, and the distinction matters because the fix happens at different stages. A mix error gets corrected at the fader. A room error gets anticipated at the production stage, before anyone plays the track anywhere.

Share

Mixing for Rooms You’ve Never Heard

Previously we wrote about what your breakdown does to the floor. This is about what the floor does to your production.

When you mix a track for DJ use, you’re not mixing for one room. You’re mixing for every room the DJ takes it to. Five rooms, fifty rooms. None of them yours.

That’s a fundamentally different target than a great bedroom mix.

The kick transient that survives room diffusion isn’t centered in the sub. It’s the attack click in the 3-5 kHz range: the snap that cuts through even when the bass frequencies are doing something unpredictable in the room’s mode structure. Protect that range. Don’t EQ it back in pursuit of a cleaner top end (it took me a long time to stop doing this). It’s doing structural work that your bedroom never makes visible.

Comb filtering is what kills vocal presence in rooms with hard parallel walls. A slight boost around 2-3 kHz gives a vocal enough cut to hold through that smearing. Too much and it becomes harsh. Too little and the vocal disappears the moment the room takes over. The difference is roughly 1-2 dB, and you can only hear it on systems that expose what your bedroom has been hiding.

The Three-System Test

Take a production you’re working on. Play it on three systems with nothing changed between them.

Phone speakers first. No sub, no bass reinforcement: just midrange and treble. If your kick vanishes here, it has no attack energy above 200 Hz. That’s the same kick that disappears in a venue where the room’s mode structure doesn’t support your sub frequency.

Car speakers next. Moderate bass, compressed midrange, a high-mid peak around 2-4 kHz from the enclosure. Notice what competes. A vocal fighting the lead synth in the car will fight it in rooms where that frequency range is already being made unpredictable by reflections.

Something with real sub last: a subwoofer, a club system, whatever you have access to. The low end reveals itself here. Frequencies that sounded distinct at home blur together at volume with real sub extension in a room that reinforces them.

One note per system. If it’s the same note three times, the mix is done. Different notes means room-dependent weaknesses the track is carrying into every venue.

Remix until the notes converge. That’s producing for the room, not your room.

Produce for the venue, not your bedroom. The DJ will play it where you’ve never mixed.

Have you ever heard your own production in a club and not recognized it? That’s a production problem, not a DJ problem.

Leave a comment