What Sonodrift Studio is
Sonodrift Studio writes and plays a complete instrumental track — drums, bass, rhythm, chords and a lead — in one of twelve styles, from a word you type in a box.
The promise in one sentence: the same seed plus the same board settings always gives you the exact same track, note for note.
There are three things to know that follow from that, things you should know before you start working with Sonodrift Studio.
- Every sound is synthesized while the track renders — every drum, string, and synth. Every sound you hear may not be unique, but it is a brand new representation of the math that made it.
- Nothing leaves your device. The track is built in your browser. There is no upload, no account, and no round trip to a server.
- A track is not a file made of musical notation and then interpreted to sound. What it is is a set of decisions made and executed while you listen, changing never more than one bar after you change any setting on the board. The seed and the positions of the controls are the track; the WAV is one printing of it. That is why a link can carry a whole track, and why the same seed reproduces it exactly.
You can access and use every style, every control on the Setup, Feel, Musical and Roles panels, the
live mixing desk, unlimited generating, and a full-quality WAV of up to 30 seconds without costing you anyting more than a bit of your time and attention.
For longer tracks, control over loudness targets, MIDI output, stems, the certificate, saved presets and A/B compare, a Studio license can be purchased.
Your first track in 60 seconds
Open Studio and leave everything alone. The board arrives ready to go: Style on Rock, a freshly-diced Seed already in the box, Length on 30s, Seamless loop ticked, auto tempo ticked.
- Press “Generate track” — at the top of “Your track”, in the right-hand column. The line underneath changes to “Composing rock…”, then counts up a percentage. A Cancel button appears next to it if you change your mind.
- Wait a few seconds. This is real synthesis, not playback. On a fast machine a 30-second track took roughly 3.5 seconds for Ambient, 4 for Lo-fi, 5 for EDM, 8–10 for Rock and 14 for Classical. A five-minute track is about ten times the work.
- Listen to it. Studio has no file player of its own — pressing Play live plays this board straight from the engine, and the WAV you download is the same music rendered to a file. Because seamless loop is on, the desk repeats the track end-to-start.
- Not your thing? Press the dice next to the Seed box. It writes a new seed and re-renders immediately, because you already have a track on screen. Keep pressing until something catches you — this is the fastest way to explore.
- Want a different genre? Pick one from Style, then press Generate track again. Only a seed change re-renders by itself; everything else waits for you.
- Look at the LIVE line under the transport while it plays. It names what the seed chose for key and tempo. The full plan — bar count, form, sections, chords and which instruments played — is in the certificate.
- Press “↓ WAV”. A 30-second file is about 7.6 MB — 44.1 kHz, 24-bit, stereo.
That is the whole loop: dice → Generate → listen. Everything else on the board is a way of narrowing what the dice is allowed to do.
How the board is laid out
There are six areas, and they are in the order you will use them.
- The stage, across the top — Play live and its transport, then Setup (Style, Seed, Length), then the per-role meters under those. Playing is the first thing on the board, and the stage holds nothing but the things you press to hear something.
- Feel — six sliders that change how the track feels without changing what it is.
- Musical — key, scale, form, chords and tempo. Every one starts on Auto.
- Roles — the mixer. One strip per part of the band.
- Your track, in the right-hand rail — Edit track… and Seamless loop, then Generate track, then every download the render produced. Studio makes files here; it does not play them.
- Keep this and Playlist, further down the rail — Copy link, Save to my playlist, and the list itself with keep playing and Shuffle… in it.
- Session, at the bottom of the rail — two A/B slots, ten starter presets, and your own presets.
On a narrow window the columns stack, and the Roles strips drop from three across to two, then to one. The board never scrolls sideways.
Setup — the Style menu
Style is the biggest decision on the board. It is not a filter over one sound: each style brings its own instruments, its own drum patterns, its own tempo range, its own list of scales and its own chord vocabulary. Changing it rebuilds most of the rest of the board.
| Style | What it sounds like | Tempo range (starts at) | The band |
|---|---|---|---|
| Pop | Electronic kit with a programmed backbeat; the chorus lifts to four-on-the-floor with offbeat open hats. Instrumental by design — a bed for a vocal or a video. | 92–132 (116) | Electronic kit · synth bass · synth-pluck stabs · warm pad · saw lead |
| Indie-pop | Pop’s organic cousin: acoustic kit with tambourine on 2 and 4, jangle guitar, more room on everything. | 100–140 (120) | Acoustic kit · electric bass · jangle guitar · organ · chime lead |
| Rock | Snare on 2 and 4, power-chord guitar in eighth-note downstrokes, bass locked to the kick, a drum fill every 8 bars. | 96–150 (124) | Acoustic kit · electric bass · rhythm guitar · organ · lead guitar |
| Punk | The rock band played faster and drier, with no keyboard layer at all and a fill every 4 bars. | 150–220 (180) | Acoustic kit · electric bass · rhythm guitar · lead guitar |
| Metal | Hotter guitars, dark minor scales, double-kick sixteenths that thicken as Energy rises, a fill every 4 bars. No keyboard layer. | 100–200 (150) | Acoustic kit · metal bass · metal rhythm guitar · metal lead |
| EDM | Kick on every beat, offbeat open hats, plucky stabs, wide supersaw pad. Its song form runs intro–build–drop–breakdown–build–drop–outro. | 120–140 (128) | Electronic kit · synth or reese bass · pluck stabs · supersaw pad · saw lead |
| House | The house signature — an open hat on every offbeat eighth — a touch of shuffle, piano and organ stabs, and no drum fills: it rides the loop. There is no lead instrument. | 118–128 (124) | Electronic kit · deep synth bass · house piano · warm pad |
| Lo-fi | The “beats to study to” sound: heavily swung sixteenths, seventh chords instead of triads, a dusty kit, and a found-sound bed — vinyl crackle, rain, a crackling fire, even a cat purring in and out — under everything, with thunder rolling somewhere outside. | 70–90 (82) | Lo-fi kit · thunder · soft bass · electric piano · pad · lead · found-sound textures |
| Reggae | The one drop: beat 1 is empty — the kick lands with the cross-stick on beat 3, once a bar, and the bass fills the hole the drummer left. Chopped guitar on every upbeat, organ threading between, and hardly any harmony at all: one chord, or two, held two bars each. Steppers and rockers feels turn up in the louder sections. | 60–90 (72) | Reggae kit · reggae bass · skank guitar · bubble organ · melodica |
| Country | The train beat — an unbroken stream of snare with the backbeat still on 2 and 4 — or a shuffle, or brushes on a ballad. Plain major harmony, a root-and-fifth upright bass, and a pedal steel that actually bends: it is the only style whose lead slides between notes rather than stepping. | 72–152 (112) | Brush or acoustic kit · upright bass · steel-string guitar · piano · fiddle or harmonica · pedal steel |
| Ambient | Beatless. Four sustained layers and nothing to keep time. This is the sound the older Ambient board is built around. | 44–96 (60) | Deep bass · warm pad · chime · bell texture |
| Classical | Beatless chamber ensemble in a longer, warmer hall than any other style. Textbook progressions — authentic cadences, the circle of fifths, the Pachelbel ground. | 66–108 (84) | Cello · string ensemble · violin, flute, clarinet or trumpet · woodwind texture |
What changes when you switch style
- The tempo slider re-tunes to the new style’s range and jumps to that style’s starting tempo.
- The Scale, Form and Chords menus are rebuilt from the new style’s own lists and reset to Auto — a scale that is legal in Metal may not exist in House.
- Key stays where you put it. It is the one Musical control a style change does not reset.
- The Swing slider snaps to the style’s own feel: 0 for most, 0.08 for House, and 0.62 for Lo-fi — the single most identifying thing about that style.
- The Roles strips are rebuilt for the new band, with that style’s default levels and every instrument menu back on Auto. That is why Punk and Metal have no keyboard strip, House has no lead strip, and Ambient and Classical have no drums.
Copy a link (or save a preset) before you go style-shopping, or your mix is gone.
Setup — the Seed and the dice
A seed is a name for a track. It is any text you like: my-first-track, tuesday, Chorus idea 3. It is not a number, it does not mean anything musically, and there is no “good” or “bad” seed — it is simply the label the engine turns into every decision it has been left free to make.
What matters is that it is exact.
- The same seed with the same board gives the same track, every time, on any machine.
- Seeds are case-sensitive and space-sensitive. Amber and amber are two different tracks — as Rock they came out in A minor at 120 BPM and G minor at 150 BPM respectively.
- Leaving the box empty falls back to the seed my-first-track.
The dice button writes a fresh random seed in the form word-word-number — for example ember-vega-317 — drawn from a list of 25 words. It is a starting point, not a magic formula. Type your own over it whenever you want a name you will remember.
The dice also re-renders. A seed is a different tune, not an adjustment, so committing one — pressing the dice, or pressing Enter or clicking away after typing — stops live playback and immediately generates the new track. That only happens once you already have a track on screen; before your first Generate, changing the seed does nothing on its own.
Every visit starts on a new random seed, so you are unlikely to land on the same default track as everybody else. Open a shared link instead and it starts on that link’s seed.
Setup — Length
Five buttons: 15s · 30s · 60s · 2 min · 5 min. It starts on 30s.
Length is not a trim. It is composed into the piece: the bar count, the tempo the seed picks, and how the form’s sections are divided all follow from it. The same seed at 15 seconds and at five minutes is recognisably the same idea — same key, same scale, same form — but it is not the same recording made longer.
- Free is 30 seconds. The 60s, 2 min and 5 min buttons carry a “Pro” badge; clicking one opens the license dialog instead of selecting it. With a license the badges vanish and the buttons select normally. This is the one Setup control that is gated.
- Short lengths constrain the tempo. At 15 seconds there is only one tempo that fits a whole number of bars in EDM (128), House (128) and Lo-fi (80), so auto tempo always lands there. At 30 seconds EDM can be 120, 128 or 136; House 120 or 128; Lo-fi 72, 80 or 88; Rock any of six values. From 60 seconds up, every style has plenty of room.
- Song forms need room. A 30-second Rock “song” squeezes intro, two verses, three choruses, a bridge and an outro into about 15 bars — roughly one to three bars each. The same form over five minutes gives each section 12 to 24 bars. If you pin a song form, give it two minutes or more.
Opening a shared link that asks for a longer length shows you 30s if you are on the free tier, rather than pretending and then quietly shortening the file.
Seamless loop
Ticked by default, and it lives in Your track, to the right of Edit track… — with the file it shapes rather than with the tune's identity. It changes how the two ends of the file are treated.
- On: the engine keeps playing two seconds past the end of the track and folds that material back over the beginning with a crossfade. The result is a file whose last moment flows into its first — no silence, no cut-off reverb tail. The live desk also repeats the track so you can hear the join.
- Off: the file is left to end where it ends, with a 30-millisecond fade at each edge so it neither clicks in nor clicks out.
The file is exactly as long either way — the extra two seconds are folded in, not appended.
Turn it on for background beds, game loops and video underscore. Turn it off when the track is meant to finish, and you want the last chord to be the last chord.
One caveat, and it is the reason auto tempo exists: a loop only sounds seamless if the loop point lands where a bar lands. Auto tempo guarantees that for the eight rhythmic styles. Pin an awkward tempo and the join falls in the middle of a bar — see Tempo.
Setup — Copy link
“Copy link” puts a link on your clipboard that carries the entire board: seed, style, length, loop setting, all six Feel sliders, every Musical pin, every role fader, every mute and every instrument choice. The button says “Copied!” for a moment; if your browser blocks clipboard access, a box appears with the link so you can copy it by hand.
Opening that link anywhere sets the board up exactly as you left it — and then waits. It does not render on arrival; you press Generate track when you are ready. Because the seed and the settings are what determine the audio, whoever opens the link gets your track, not an approximation of it.
This is the cheapest way to:
- save a track you like — paste the link into your notes; saving presets needs a license, a link does not;
- send a track to someone without sending a file;
- come back to a board tomorrow, on another machine.
A link made on a free board carries free settings. A link made with a license that asks for a five-minute length or a loudness target shows a free user the honest 30-second, Natural version rather than a setting they cannot render.
Feel — the six sliders
The Feel panel sits between Setup and Musical and holds six sliders: Energy, Motion, Swing, Space, Brightness and Stereo width. They are the fastest way to change how a track feels without changing what it is.
Four things are true of all six.
- They are all free. Nothing in the Feel panel is behind the license.
- They all really change the audio. None is a cosmetic knob; each is fed into the synthesis.
- They are all written into the certificate. Anyone re-running the engine with your seed and your six values gets a bit-identical file back.
- The downloaded WAV only changes when you press Generate track. Moving a slider afterwards does not alter the file you already made.
| Slider | Range | Ships at | What it changes | How big it is |
|---|---|---|---|---|
| Energy | 0 – 1 | 0.5 | How hard the band plays: drum velocities, how many optional hits land, how many melody notes sound, and whether thin sections keep their bass and rhythm parts | Modest — about 1 to 2 dB across the whole sweep |
| Motion | 0 – 1 | 0.35 | A slow rise and fall in the loudness of the pad, lead and texture parts, one cycle every 12–20 bars | Subtle; clear on Ambient and Classical over long takes, nearly invisible on the band styles |
| Swing | 0 – 1 | the style’s own value | Pushes every off-beat sixteenth of the drum pattern later | Obvious where there is a kit; nothing at all on Ambient and Classical |
| Space | 0 – 1 | 0.3 | How much of the band is fed to the reverb | Audible in every style; biggest on Ambient, Classical and Lo-fi |
| Brightness | 0 – 1 | 0.5 | The harmonic tilt of every pitched instrument; drums are untouched | Gentle downward; upward is bounded per instrument — deep-stack voices cap their own response (2.8.0) |
| Stereo width | 0 – 1.5 | 1.0 | How far the stereo image spreads; 0 is exact mono | Obvious at the extremes |
Stereo width is the only one that goes past 1 — 1.0 is its neutral center, not its maximum. The six are also not equally dramatic: Brightness, Space, Stereo width and Swing change a track obviously, Energy is real but modest, and Motion is genuinely subtle and only on some styles. While the live desk is running, they arrive at different times too — see what lands instantly.
Energy — how hard the band plays
0 to 1, ships at 0.5, and 0.5 is an exact neutral: you get the style’s own written arrangement, untouched.
Every style writes an energy level into each section of its form — a rock intro at 0.4, a verse at 0.6, a chorus at 0.9. The slider shifts all of those up or down together, by up to about a third at the extremes. That shift reaches every instrument:
- Drum hits the pattern marks as optional become half as likely at 0 and half again as likely at 1. Anchor hits — a rock snare on 2 and 4 — never disappear.
- Every drum hit is played harder or softer, over a range of about 3 dB end to end.
- Melodies rest less and play more of the notes in each bar; chord and bass parts are struck harder.
- The biggest effect: thin sections can lose instruments entirely. A section that falls low enough drops its bass, and the rhythm part goes with it.
How big it actually is
Across ten seeds and six styles, sweeping Energy from 0 to 1 moved the finished mix by roughly 0.9 to 2.0 dB — rock and pop about 1.5 dB, EDM 1.2, classical 1.0, ambient and lo-fi up to 2.0. Note counts move by only a few per cent. Do not treat Energy as a loudness control; the Level faders in the Roles panel move much further.
Where a role actually drops out at low settings, the lower half of the slider does about twice as much as the upper half — measured at roughly 2 to 3 times on rock, EDM, lo-fi and ambient. On the beatless styles and on seeds that drew a loop form, it is close to linear.
The top of the slider does less than you would expect, and how much less depends on the style. The loudest section each style writes runs from 0.7 (Ambient) and 0.75 (Lo-fi) through 0.9 (Classical) and 0.95 (Rock, Pop, Indie-pop) up to a flat maximum in EDM, House, Metal and Punk — and in those four the peak sections are already at the ceiling with Energy at its default, so the whole upper half of the slider does nothing to them at all.
Pick Rock, keep the seed, generate at Energy 0.5, then again at Energy 0. Listen to the intro, not the choruses. At 0.5 the intro has drums, organ, bass and rhythm guitar. At 0 it is drums and organ only — the bass and the guitar have gone. On some seeds the outro empties out the same way. Lo-fi behaves identically: at 0 its intro, outro and breakdown all lose their bass.
The honest catch. This only happens if your seed drew a form with an intro and an outro — for Rock that is the song form, not the two-part loop form. On a loop-form seed no instrument ever drops out at any Energy setting, and you get density and velocity changes only. If you moved Energy and heard nothing, check the form: the certificate’s section list is where an intro shows up.
Motion — a slow swell over long takes
0 to 1, ships at 0.35.
The seed draws a slow wave that completes one cycle every 12 to 20 bars, and Motion decides how deeply that wave raises and lowers the loudness of the notes played by the pad, lead and texture parts. At maximum the swing is about a third of note loudness either way. Drums, bass and the rhythm part are never touched.
On Ambient and Classical, pad, lead and texture are the piece, so the swell is the piece breathing. On the band styles those roles are a small slice of a mix dominated by drums, bass and guitars, so the swell is applied but largely masked. Measured as the spread between the loudest and quietest bar of a two-minute take:
| Style (2 min) | Motion 0 | Motion 1 | Motion added |
|---|---|---|---|
| Ambient | 2.65 dB | 5.53 dB | +2.9 dB |
| Classical | 2.02 dB | 3.88 dB | +1.9 dB |
| Rock | 4.22 dB | 4.80 dB | +0.6 dB |
On rock the swell is genuinely there — isolate the pad and lead and their spread nearly doubles, from about 5.5 dB to 10.5 dB. It is simply buried under the kit and the guitars.
It needs length. A 12-to-20-bar cycle at Ambient’s slow tempo is roughly 50 to 90 seconds. A 30-second Ambient take is about seven bars, so you hear less than half a cycle and it reads as a one-directional drift rather than a swell. On the same seed, Motion added 2.9 dB of range at two minutes but only 1.1 dB at 30 seconds. Since free tracks cap at 30 seconds, Motion mostly pays off with a license and the longer lengths.
Ambient, two or five minutes, seamless loop on. Generate at Motion 0, then at Motion 1. Play both all the way through and listen to the pad: at 1 it rises and falls under you, at 0 it sits still. Do not compare the first ten seconds — there is nothing to hear there yet.
The honest summary: Motion is the subtlest of the six. On Ambient and Classical at length it is a real, musical improvement. On Rock, Pop, EDM, House, Punk and Metal you should not expect to hear it.
Swing — the drums only, and only where there are drums
0 to 1. It has no fixed default: it snaps to whatever the style you picked authored, every time you change style.
| Style | Snaps to | What that means at the style’s preferred tempo |
|---|---|---|
| Lo-fi | 0.62 | Off-beat sixteenths land about 38 ms late — the whole identity of the style |
| House | 0.08 | About 3 ms late — honestly, below what most listeners will hear |
| Pop, Indie-pop, Rock, Punk, Metal, EDM | 0 | Dead straight, as written |
| Ambient, Classical | 0 | These styles have no drum kit; the slider does nothing at all |
It delays every odd sixteenth-note step of the drum pattern, up to about a third of a sixteenth at maximum — around 41 ms at rock tempo, 62 ms at lo-fi tempo. It moves the drums and nothing else. Bass, guitars, keys and lead stay on the grid, so at high settings you get a shuffling kit under a straight band.
On Ambient and Classical it is genuinely inert, not merely subtle: a ten-second Ambient render at Swing 0 and one at Swing 1 are byte-for-byte the same file. There is no drum pattern for the control to displace. The slider is there because the panel is shared; ignore it on those two styles.
Pick Lo-fi, generate once as it comes with Swing at 0.62, then drag Swing to 0 and generate again. The second take is the same music with the shuffle removed — it goes from a head-nod to a metronome. Then try the reverse on Rock: push Swing to 0.4–0.5 and listen to the hi-hats fall behind the snare.
Two catches. Changing style resets it — dial in a swing you like on rock, switch to pop, and the slider snaps to pop’s authored 0. Copy a link or save a preset first; both restore the exact value. And House ships at about 3 ms of shuffle, which is real in the score but under most people’s threshold; if you want House to actually swing, push it up yourself — 0.15 to 0.25 is a usable range.
Space — the reverb
0 to 1, ships at 0.3. If you tried this slider on an older build and concluded it did nothing, try it again: a mis-ported setting used to run the reverb about 15 dB too quiet at every position, so sweeping it end to end changed a mix by hundredths of a decibel. That is fixed. The reverb itself was always well tuned; it just arrived far too quiet.
Space is the master control on a single shared reverb. Each style decides which parts are sent to it and how much: pads and leads are wettest, drums close to dry, and the bass is completely dry in every style except Classical and Lo-fi, so the low end never muds. Classical plays in its own larger hall — about 2.4 seconds of tail against about 1.3 for the default room.
| Style | Reverb at the default 0.3 | Reverb at maximum | Change to the whole mix, 0 → 1 |
|---|---|---|---|
| Ambient | −12.0 dB | −1.6 dB | +2.65 dB |
| Classical | −12.5 | −2.0 | +1.91 |
| Lo-fi | −13.7 | −3.3 | +0.58 |
| Indie-pop | −16.0 | −5.6 | +0.95 |
| House | −17.2 | −6.7 | +0.59 |
| Pop | −19.4 | −9.0 | +0.53 |
| EDM | −20.9 | −10.4 | +0.35 |
| Rock | −21.7 | −11.2 | +0.41 |
| Metal | −24.9 | −14.4 | +0.02 |
| Punk | −26.6 | −16.1 | +0.00 |
That ladder is deliberate: Ambient and Classical are rooms, Punk and Metal stay raw. In every style the slider moves the reverb by at least 10 dB from the default to the maximum, so it always does something — but what that does to the finished mix varies enormously. On Ambient, Classical and Lo-fi, Space is a headline control and dragging it moves the track to a different room. On Punk and Metal you will hear a tail on the snare and the guitar if you listen for it, and the mix as a whole barely changes, because those styles send almost nothing to the reverb by design. Space cannot put reverb on a part the style keeps dry.
Ambient or Classical, Space at 0, generate; then Space at 1.0, generate. Listen to what happens after a chord stops — at 0 the note simply ends, at 1 it hangs. On Rock, do the same and listen only to the snare.
One trap. Raising Space also raises the overall level, by up to 2.65 dB on Ambient. Louder always sounds better in a quick comparison, so turn your monitoring down a little when you audition the wetter version, or you will pick it for the wrong reason.
With a license, the stems download includes the reverb return as its own file — the clearest possible way to hear exactly what Space is adding.
Brightness — the biggest control on the panel, and the one to respect
0 to 1, ships at 0.5, and 0.5 is exactly neutral — every instrument renders precisely as it was designed and ear-checked. Move it at all and you are off the tuned path in one direction or the other.
It tilts the harmonics of every pitched instrument: below 0.5 each successive harmonic is reduced, above 0.5 each is boosted. Drums are completely unaffected — a drums-only render at 0 and at 1 is byte-for-byte identical. So on a busy band track, Brightness moves the guitars, keys, bass and lead while the kit stays put.
It is not symmetrical, and it is not the same size on every style. The boost compounds harmonic by harmonic, so instruments built from deep harmonic stacks — the electric guitars, and Classical’s strings, piano, flute and clarinet — react far more than a simple synth pad.
The screech is gone (2.8.0). The deep-stack instruments whose boost compounded into an ear-piercing wall — String ensemble, both rhythm guitars, Clean electric, Nylon, Jangle — now scale the slider's sweep onto their own usable range (String ensemble and the rhythm guitars top out at an internal 0.6, Clean electric 0.65, Nylon 0.75, Jangle 0.8). Full-right is the brightest usable setting on every instrument, not a distortion effect: the slider's whole travel stays expressive, and what changes per instrument is how far the top actually tilts.
Practical ranges:
- All styles: the whole slider is usable. Below 0.5 everything darkens gently — under 2 dB of tilt from 0.5 down to 0.3.
- Rock, Metal: the capped rhythm guitars hold their tone at the top while the uncapped leads still open up — the upper third now brightens the topline over a stable bed instead of clipping the whole mix.
- Ambient: the mildest of all — end to end, about 3.5 dB of tilt and 1 dB of level. If you want Ambient to change character, Space and Stereo width will do more for you.
Rock, one seed. Generate at 0.45, then 0.5, then 0.55, and listen to the pick attack on the rhythm guitar and the top of the organ. Those three settings cover the entire musically useful range on that style — worth knowing before you spend time up at 0.8 wondering why it sounds broken.
Stereo width — from mono to very wide
0 to 1.5 — the only slider that goes past 1. Ships at 1.0, which is neutral: the instrument positions the style wrote, unmodified.
It scales the difference between the left and right channels. At 0 that difference is removed entirely and the file is exact mono — the two channels are sample-for-sample identical. Above 1 the difference is exaggerated and the image opens out past where the style placed things.
How wide each style already is, measured as side content relative to center content:
| Style | at 0.5 | at 1.0 (neutral) | at 1.5 |
|---|---|---|---|
| EDM | −19.3 dB | −13.3 dB | −9.7 dB |
| Rock | −14.5 | −8.5 | −5.0 |
| Ambient | −9.7 | −3.7 | −0.2 |
| Classical | −5.9 | +0.1 | +3.6 |
Classical and Ambient are already very wide at neutral — Classical’s two channels are essentially unrelated at 1.0. EDM and Rock are much more centered and have real room to open up.
The honest caution. Pushing a style that is already wide up to 1.5 produces content that partially cancels when the track is played in mono — a phone speaker, a laptop, a club system, a Bluetooth speaker on a shelf. At 1.5 the Classical take measures negative channel correlation, which is the signature of exactly that problem; Ambient at 1.5 is right at the edge. EDM at 1.5 is still comfortably safe.
Width also changes the level — a Classical take measures −20.4 dB at width 0 and −15.4 dB at 1.5 — so, as with Space, level-match before you judge.
When you have a take you like, drag Stereo width to 0 and listen once in mono. If the lead thins out or the pad half-vanishes, you were too wide: bring it back toward 1.0. Then, if you want more width, add it on the tighter styles — EDM, Rock, Pop — where there is headroom, rather than on Ambient and Classical where there is not.
Which sliders will reward you
Ranked by how much you will actually hear, so you can spend your time in the right place.
- Brightness — the largest and most immediate change on the panel. The whole slider is usable on every style: instruments that used to screech at the top now cap their own response (2.8.0).
- Space — clearly audible in every style. Transformative on Ambient, Classical and Lo-fi; a real but modest tail on Punk and Metal, which are dry on purpose.
- Stereo width — obvious at the extremes, and the mono check at 0 is a genuinely useful habit. Neutral is already well judged.
- Swing — the entire personality of Lo-fi, and a strong effect on any style with a kit. Completely inert on Ambient and Classical, and easy to lose because it re-snaps whenever you change style.
- Energy — real but modest, and the interesting half is 0 to 0.5. Its best trick is not level at all: it is dropping the bass and rhythm guitar out of thin sections, and that only happens on seeds that drew an intro-and-outro form.
- Motion — the subtlest by a distance. Worth setting on Ambient and Classical at two minutes or longer. On the band styles, do not expect to hear it.
A sane first pass on any track: leave Energy and Motion where they ship, let Swing stay wherever the style put it, and spend your time on Space, Brightness and Stereo width. Then, if the track is Ambient or Classical and longer than a minute, come back and push Motion up; if it is a band style with an intro, come back and pull Energy down and listen to what leaves the room.
Musical — what “Auto” means
Every menu in the Musical panel starts on Auto, and the panel’s own subtitle is the whole idea: Auto lets the seed decide, inside the style’s rules.
Read that literally. Auto is not “random each time you press Generate”, and it is not “the same safe default every time”. It means the seed makes the choice, from the list the style allows. So:
- The choice is stable — same seed, same choice, forever.
- The choice is legal — the seed can only pick from things that style actually plays, so an Auto board is never out of style and never out of key.
- The choice is visible — the LIVE line names the key and tempo it picked, and the plan in the certificate names all of it.
Pinning a menu takes that one decision away from the seed and hands it to you. Everything still on Auto keeps varying with the seed. This is the intended way to work: dice until something is close, pin the one thing you want to keep, then dice again to vary everything else around it. A pin is only ever a constraint on a decision, never a new sound — the style still decides how to play in the key or the progression you chose.
Musical — Key and Scale
Key offers Auto (seed decides) plus the twelve pitch classes: C, C♯, D, D♯, E, F, F♯, G, G♯, A, A♯, B. They are written as sharps only — there is no “E♭” entry, so pick D♯.
- On Auto, all twelve are equally likely — over 240 seeds each key came up between 12 and 29 times.
- The key comes from the seed alone. The same seed gives the same key in every style and at every length. That makes the key a stable handle: change style to hear the same key played by a different band.
- Key is the one Musical menu that survives a style change. Pin C and it stays C as you audition all twelve styles.
- Pin it when you are writing to something — an existing vocal, a video with a note in it, another Sonodrift track you want to sit beside.
Scale offers Auto plus that style’s own scales, listed most-likely-first — the top entry is the one Auto picks most often.
| Style | Scale menu (most likely first) |
|---|---|
| Pop | Major (ionian) · Mixolydian · Minor (aeolian) · Dorian |
| Indie-pop | Major · Mixolydian · Dorian · Lydian |
| Rock | Mixolydian · Major · Minor · Dorian |
| Punk | Major · Mixolydian · Minor · Major pentatonic |
| Metal | Minor · Phrygian · Minor pentatonic · Harmonic minor |
| EDM | Minor · Major · Dorian · Minor pentatonic |
| House | Dorian · Minor · Major · Minor pentatonic |
| Lo-fi | Dorian · Minor · Major |
| Ambient | Lydian · Dorian · Major · Minor · Mixolydian |
| Classical | Major · Minor · Harmonic minor · Lydian |
Scale is the strongest mood control on the board and the cheapest thing to experiment with. Same seed, same style, same everything: switch Rock from Mixolydian to Minor (aeolian) and the whole track darkens without a single note moving out of place, because the chords follow the scale.
A scale pinned in one style drops back to Auto if you switch to a style that does not offer it.
Musical — Form
Form is the shape of the arrangement: how many sections there are and what each one does. The menu is Auto plus that style’s templates.
- Eight styles offer loop and song. Loop is two halves that repeat; song is a real arrangement — for Rock: intro, verse, chorus, verse, chorus, bridge, chorus, outro. Pop’s song adds a prechorus; Punk and Metal swap the bridge for a breakdown; EDM’s song is intro, build, drop, breakdown, build, drop, outro; House’s is intro, verse, chorus, breakdown, chorus, outro.
- Ambient and Classical offer loop and arc instead. Ambient’s arc is intro → development → outro; Classical’s is intro → development → climax → outro.
- Lo-fi has three, and its menu entries keep the style’s own prefix: lofi-loop (two halves), lofi-study (intro, then A–B–A) and lofi-tape (a full intro, verse, chorus, breakdown, chorus, outro tape).
On Auto the loop template is the most likely: roughly 60% loop against 40% song or arc for eight styles, and for Lo-fi about 44% lofi-loop, 33% lofi-study, 22% lofi-tape.
Form interacts strongly with Length. Sections are handed out in proportion, with a one-bar minimum, so a long form on a short track gets compressed hard — a 30-second Rock song gives the intro one bar and the bridge one bar. Pin loop for background music and short loops; pin song or arc when you have given the track two minutes or more and want it to go somewhere.
Like Key, Form is decided by the seed alone: changing the length does not change which form a seed picks, only how many bars each section gets.
Musical — Chords
The Chords menu is Auto (seed draws) plus that style’s progressions, named the way musicians name them.
| Style | Chords menu entries |
|---|---|
| Pop | I-V-vi-IV · vi-IV-I-V · I-vi-IV-V · I-IV-vi-V |
| Indie-pop | I-V-vi-IV · IV-I-V-vi · I-IV-V-IV · vi-IV-I-V |
| Rock | I-bVII-IV-I · I-V-vi-IV · vi-IV-I-V · I-IV-V-IV |
| Punk | I-IV-V · I-V-vi-IV · I-bVII-IV · vi-IV-I-V |
| Metal | i-bVI-bVII-i · i-bVII-bVI-v · i-iv-i-v · i-i-bVI-bVII |
| EDM | vi-IV-I-V · i-VI-III-VII · i-VII-VI-VII · I-V-vi-IV |
| House | i-IV · i-VII-IV-i · i-IV-V-IV · vi-IV-I-V |
| Lo-fi | lofi-i7-IV7 · lofi-ii-V-i · lofi-i-VII-VI-V · lofi-i7-sus-IV7 · lofi-i-III-VI-V |
| Ambient | 0 · 1 · 2 · 3 · 4 · 5 · 6 · 7 |
| Classical | cadence · circle · half · predominant · pachelbel |
This is the one Auto that keeps making new decisions all the way through the track. Every four bars — or however long the current progression is — an unpinned board draws a fresh progression. A 60-second Rock track on Auto drew seven progressions in a row: I-♭VII-IV-I, then vi-IV-I-V, then I-V-vi-IV twice, then I-IV-V-IV, and so on. That is what stops a long track being the same five seconds over and over.
Pinning a progression locks it for the entire track. The same 60-second Rock track with I-V-vi-IV pinned plays I-V-vi-IV for twenty-five bars in a row. That is exactly what you want for a house or lo-fi vamp, and exactly what you do not want for a five-minute piece you intend to sit and listen to. So: pin for hypnotic, leave on Auto for developing.
Two notes on the names. Ambient’s eight progressions are numbered only because they have no common names — audition them; they are all equally likely on Auto. Lo-fi’s entries keep their prefix and describe seventh-chord harmony: lofi-ii-V-i is the jazz cadence, lofi-i7-IV7 the two-chord vamp that defines the genre.
Musical — Mood and Detail
Mood leans the seed's own choices without ever overriding yours. Pick Dark and the seed favours the style's minor-leaning scales and falling melodic shapes; pick Uplifting and it favours the bright ones, a touch faster; Calm sits the tempo low and keeps the tune stepping rather than leaping. It is a lean, not a switch: every mood draws only from the scales the style already owns, so Metal stays Metal in any mood — and anything you pin (a Key, a Scale, the tempo) always wins outright. The Auto labels on Key and Scale re-resolve when you change Mood, so what they name is always what will play.
Detail is how elaborate the playing is, at the same tempo and loudness. Simple leans every instrument toward its plainest figures — held chords, root notes, fewer ghost hits on the kit. Intricate leans toward the busy end: walking lines, sixteenth figures, sequenced and inverted melodic phrases, a livelier optional layer on the drums. The backbone never changes — anchors like the backbeat and the four-on-the-floor are untouchable — and Auto is the balanced middle the styles were tuned at.
Both are recorded in the certificate like every other control, both travel in share links and presets, and both can be marked for Shuffle so a new tune re-rolls them.
Musical — Tempo and auto tempo
The tempo slider is bounded by the current style’s range and starts at that style’s preferred tempo. The auto tempo box next to it is ticked by default, and while it is ticked the read-out says auto instead of a number. Dragging the slider unticks auto tempo automatically — there is no separate step. Tick the box again to hand tempo back to the seed.
What “auto” actually does depends on the style
- For the ten rhythmic styles — Pop, Indie-pop, Rock, Punk, Metal, EDM, House, Lo-fi, Reggae, Country — auto tempo solves for an exact fit: from the tempos in that style’s range it picks one that makes your chosen length come out as a whole number of bars. A 30-second Rock loop is exactly 12, 14 or 15 bars and ends precisely where it began. Measured across 24 seeds at all five lengths, it was exact every single time. This is what makes seamless loop actually seamless — no trailing half-beat, no drum flam at the join.
- For Ambient and Classical there is no exact fit, because there is no beat to protect. Auto draws a tempo from the range, rounds the bar count up, and the composed material simply runs past the end of the file and is cut there. On one test seed, Ambient at 30 seconds chose 94 BPM and 12 bars — 30.64 seconds of music in a 30-second file; Classical chose 106 BPM and 14 bars, or 31.70 seconds. Because both styles are beatless this is rarely audible, but it is not the same guarantee the rhythmic styles get.
Because it is solving an equation, auto tempo can land on a fraction of a beat per minute on long tracks — a five-minute Rock track might come out at 122.4 or 97.6 BPM. That is not a rounding error; it is the tempo at which 153 bars take exactly 300 seconds. Tempos you pin yourself are always whole numbers.
Pinning a tempo costs you the exact fit. Rock at 30 seconds with 140 BPM pinned needs 18 bars, which run 30.9 seconds — so the file stops just under a bar short, and with seamless loop on the join lands in the middle of a bar where you will hear it. If you need a specific tempo and a clean loop, pick a length the tempo divides evenly. At 140 BPM a bar is 1.714 seconds, so 60 seconds is not clean either; 15 or 30 seconds at 128 BPM is.
One more behavior worth knowing: if you have pinned a tempo and then switch style, the pin stays on but the number jumps to the new style’s preferred tempo — because 180 BPM means something in Punk and nothing in Ambient. Glance at the read-out after a style change.
Roles: your band, one strip per part
The Roles panel is Studio’s mixer. Each strip is one part of the band — not one instrument, but a job in the arrangement.
| Role | The job it does |
|---|---|
| Drums | The kit: kick, snare, hats, ride, crash — the whole groove on one strip |
| Weather | Ambient and Lo-fi: rolling thunder, timed in seconds rather than in bars |
| Bass | The low line under the chords |
| Rhythm | The chordal engine room — power-chord guitar, piano stabs, plucked chords |
| Harmony | Sustained chords: pads, organ, strings |
| Lead | The topline melody |
| Texture | Atmosphere — vinyl crackle, rain, fire, bells, air. Always centred, and never rests. |
Strips are always shown in that order, and each style fields a different band, so the panel changes shape when you change Style. Punk and Metal have no Harmony strip — nothing holds pads under a wall of guitars. House has no Lead: its topline is the piano, which lives on Rhythm. Ambient and Classical have no Drums at all. Lo-fi is the widest board at seven strips — it adds the found-sound bed on Texture and carries Weather; Ambient fields five.
Weather is the odd one, and deliberately so. It is the only layer in Studio that is not timed in bars: thunder arrives on wall-clock time, each strike one whole roll — a few climbing swells into a loudest peak, then a tail of ever-smaller swells receding into the distance. No two rolls are alike. How near the storm sounds, how deep it rumbles, how much it crackles, how much it echoes, how far it drifts across the stereo field, and how many swells rise and fall are all drawn from the seed, so a long session gives you a storm that moves rather than one sound on repeat — and the same seed always gives the same storm back.
Weather is on Ambient and Lo-fi. Its instrument menu holds four different storms, and picking one changes the weather, not just the tone — each brings its own cadence and roll length. Summer thunder is distant and smooth, rolling slowly from one side of the field to the other, and comes about once a minute. Canyon thunder is long, with big irregular echoes arriving from different directions. Storm is close, powerful and frequent, and travels least of the four. Dry thunder is the short, sharp, close one — the most startling of the four, which is why it is also the rarest. Pull the fader up for a storm, down to nothing for clear weather, and pair any of them with Rain on Texture — that combination is what the layer was built for. No two strikes are alike: how long the roll runs, how its swells are spaced, how loud each one is, and how far it drifts across the field are all drawn fresh for every strike. In a seamless loop every sound recurs with the length of the loop, so at 15 seconds the thunder stops being a series of events and becomes one continuous rumble that rises and subsides instead.
The number after each name below is that strip’s starting Level — the balance the style ships with, and a decent reference point to mix against.
| Style | Strips, in panel order, with their default Level |
|---|---|
| Pop | Drums 1 · Bass 0.9 · Rhythm 0.65 · Harmony 0.55 · Lead 0.7 |
| Indie-pop | Drums 0.95 · Bass 0.9 · Rhythm 0.85 · Harmony 0.5 · Lead 0.6 |
| Rock | Drums 1 · Bass 1 · Rhythm 0.9 · Harmony 0.5 · Lead 0.7 |
| Punk | Drums 1 · Bass 1 · Rhythm 0.95 · Lead 0.55 |
| Metal | Drums 1 · Bass 1 · Rhythm 1 · Lead 0.55 |
| EDM | Drums 1 · Bass 1 · Rhythm 0.7 · Harmony 0.7 · Lead 0.75 |
| House | Drums 1 · Bass 1 · Rhythm 0.75 · Harmony 0.6 |
| Lo-fi | Drums 1 · Weather 0.3 · Bass 0.48 · Rhythm 0.57 · Harmony 0.4 · Lead 1 · Texture 0.34 |
| Ambient | Weather 0.4 · Bass 1 · Harmony 1 · Lead 1 · Texture 1 |
| Classical | Bass 1 · Harmony 1 · Lead 0.85 · Texture 0.45 |
Lo-fi’s numbers look lopsided and are not a mistake: its instruments are voiced at very different levels, and these fader positions are what put them where the style wants them. Its Lead already sits at the top of its fader and still lands roughly 8 dB under the chord piano — that style’s topline is meant to be background. If you want it forward, turn the other roles down; there is nothing above 1.
Each strip also carries a fixed amount of reverb send, chosen by the style — drums nearly dry, pads wettest. Those sends are not exposed as controls; the Space slider scales the whole room. Nothing in the Roles panel costs money: faders, instrument menus and mute are free in every style.
The instrument menus
Under each strip name is that role’s instrument menu. It always opens on Auto, and lists the instruments this style allows in this role. Every strip in every style has at least two, so every strip has a menu.
The Auto line names the instrument the seed has actually drawn — on Rock it reads Auto (Rock organ) or Auto (Grand piano) rather than a generic “seed picks”, so you can see the whole band the seed has assembled before you render anything. It re-reads as you type a seed, roll the dice or change style. Pinning an instrument does not change that line: it keeps telling you where Auto would go back to.
Auto means the seed draws the instrument, using odds the style sets. Two things follow.
- The odds are not even. Rock’s Harmony lands on Organ about three times in four and Grand piano about one in four — so re-rolling the seed genuinely is a way to shop for a different sound.
- Some instruments are never drawn automatically. They are in the menu, they sound fine, and Auto will not pick them however many seeds you try. Rock’s Clean electric rhythm guitar is one; Pop’s Jangle guitar is another. If you want those, you have to pick them.
Picking an instrument changes only the sound of that part. It does not rewrite the part and it does not disturb anything else: with the same seed, pinning an instrument leaves every note in every role — including the pinned one — exactly as it was. That makes the menu a safe comparison: hold the tune still, change the timbre. To go back, choose Auto again.
“Auto can land on” lists the odds; “only if you pick it” lists instruments Auto will never draw.
| Style | Strip | Auto can land on | Only if you pick it |
|---|---|---|---|
| Pop | Drums | Electronic kit | Acoustic kit |
| Pop | Bass | Synth bass | Sub bass |
| Pop | Rhythm | Synth pluck (3 in 4), Grand piano (1 in 4) | Electric piano, Jangle guitar |
| Pop | Harmony | Warm synth pad | Supersaw, Glass pad |
| Pop | Lead | Saw lead | Square lead, Chime |
| Indie-pop | Drums | Acoustic kit | Electronic kit |
| Indie-pop | Bass | Electric bass | Sub bass |
| Indie-pop | Rhythm | Jangle guitar (3 in 4), Nylon guitar (1 in 4) | Clean electric, Electric piano |
| Indie-pop | Harmony | Organ | Warm synth pad, Electric piano |
| Indie-pop | Lead | Chime | Bell, Saw lead |
| Rock | Drums | Acoustic kit | Electronic kit |
| Rock | Bass | Electric bass | Metal bass |
| Rock | Rhythm | Rhythm guitar | Metal rhythm guitar, Clean electric |
| Rock | Harmony | Organ (3 in 4), Grand piano (1 in 4) | Electric piano, Warm synth pad |
| Rock | Lead | Lead guitar | Metal lead guitar |
| Punk | Drums | Acoustic kit | Electronic kit |
| Punk | Bass | Electric bass (3 in 4), Metal bass (1 in 4) | — |
| Punk | Rhythm | Rhythm guitar (3 in 4), Metal rhythm guitar (1 in 4) | Clean electric |
| Punk | Lead | Lead guitar | Metal lead guitar |
| Metal | Drums | Acoustic kit | Electronic kit |
| Metal | Bass | Metal bass | Electric bass |
| Metal | Rhythm | Metal rhythm guitar (3 in 4), Rhythm guitar (1 in 4) | — |
| Metal | Lead | Metal lead guitar (3 in 4), Lead guitar (1 in 4) | — |
| EDM | Drums | Electronic kit | Acoustic kit |
| EDM | Bass | Synth bass (3 in 4), Reese bass (1 in 4) | Sub bass |
| EDM | Rhythm | Synth pluck | Supersaw |
| EDM | Harmony | Supersaw (3 in 4), Warm synth pad (1 in 4) | Glass pad |
| EDM | Lead | Saw lead | Square lead, Supersaw |
| House | Drums | Electronic kit | Acoustic kit |
| House | Bass | Synth bass | Sub bass, Reese bass |
| House | Rhythm | House piano (3 in 5), Grand piano (2 in 5) | Electric piano, Organ |
| House | Harmony | Warm synth pad | Supersaw, Glass pad |
| Lo-fi | Drums | The lo-fi kit | Acoustic kit |
| Lo-fi | Weather | Summer thunder (3 in 7), Canyon thunder (2 in 7), Storm (1 in 7), Dry thunder (1 in 7) | — |
| Lo-fi | Bass | Soft bass (3 in 5), Sub bass (1 in 5), Deep bass (1 in 5) | — |
| Lo-fi | Rhythm | Electric piano (4 in 9), Grand piano (2 in 9), Nylon guitar (2 in 9), House piano (1 in 9) | — |
| Lo-fi | Harmony | Warm synth pad (1 in 2), Glass pad (1 in 3), String ensemble (1 in 6) | — |
| Lo-fi | Lead | Electric piano (3 in 8), Clean electric (1 in 4), Flute (1 in 4), Synth pluck (1 in 8) | — |
| Lo-fi | Texture | Vinyl crackle (4 in 10); Bell, Rain, Crackling campfire, Cat purring, Pencil on paper, Clock ticking (1 in 10 each) | — |
| Ambient | Weather | Summer thunder (3 in 7), Canyon thunder (2 in 7), Storm (1 in 7), Dry thunder (1 in 7) | — |
| Ambient | Bass | Deep bass | Sub bass |
| Ambient | Harmony | Warm pad (3 in 4), String ensemble (1 in 4) | Warm synth pad, Glass pad |
| Ambient | Lead | Chime (3 in 4), Flute (1 in 4) | Bell, Electric piano |
| Ambient | Texture | Bell (3 in 4), Marimba (1 in 4) | Glass pad, Warm pad, Rain |
| Classical | Bass | Cello | Nylon guitar |
| Classical | Harmony | String ensemble (3 in 4), Grand piano (1 in 4) | — |
| Classical | Lead | Violin (1 in 2), Flute (1 in 4), Clarinet (1 in 8), Trumpet (1 in 8) | — |
| Classical | Texture | Flute (3 in 5), Clarinet (1 in 5), Marimba (1 in 5) | — |
Three to try. Rock with Rhythm on Clean electric and Harmony on Grand piano — the piano-led rock board Auto will never hand you. Ambient with Harmony on Glass pad. Classical with Lead on Trumpet, which Auto reaches about one seed in eight.
Level and Mute
Level
The Level fader sets how loud that role sits in the mix. It runs 0 to 1 in steps of 0.01, and the number beside it is the exact value. It is a plain linear gain, not a decibel fader: 0.5 is exactly half the amplitude of 1, about 6 dB down, and 0.25 is half of that again. There is no headroom above 1, so mixing is done by pulling other roles down.
The fader is the real mix, not a preview control. Whatever it says goes into the live sound, the downloaded WAV, the certificate that documents the track, and the per-role stems if you export them. It does not change what is played — a role you pull to 0.2 is still playing every note it was playing.
Working method. Generate once, press Play live, then mix by ear while it loops — fader moves are heard immediately. When it sits right, press Generate again; the file is rendered from the board as it stands.
Mute
Each strip has a checkbox in its top-right corner. Checked means the role plays. Uncheck it to take that role out. Mute is level times zero, and that is the whole mechanism, which is why it is honest:
- The role goes silent in the live preview, immediately.
- It goes silent in the downloaded WAV — exactly silent, not “very quiet”.
- The certificate records it as zero, so the document matches the file. A board with a role muted still renders, still certifies, and still verifies.
- That role gets no stem. Export stems with the drums muted and there is no drums file — not a silent one to sort through.
The fader stays where you left it while the role is muted, so unmuting puts the part back at the level you had. The strip dims so you can see at a glance who is out, but its menu and fader still work — you can line up an instrument on a muted strip and unmute into it.
There is no solo button. With mute landing instantly during live playback, muting the others is the practical substitute.
Two things mute does not do:
- It does not remove the part from a MIDI export. MIDI is the score, not the mix, so a muted role still exports its track in full. Delete it in your DAW.
- It does not change what the meters show — they read the parts before the faders.
Capture behaves sensibly. A share link, a preset or an A/B slot taken while a role is muted comes back muted, because the zero is the mute. One consequence: on recall, the muted strip’s fader sits at that style’s default rather than where you had it. Unmute and adjust from there. Dragging a fader all the way to 0 by hand is the same thing, and recalls as muted too.
Try it. Rock, then mute Lead and Drums for a rhythm-section bed. Or Lo-fi with everything muted but Drums and Texture — a crackly loop to play over.
Play live: mixing while it plays
Play live starts a stream of the current board and loops it until you stop — or, with keep playing ticked, hands over to a different tune each time this one has had its turn. You do not need to Generate first — press it on a fresh page and the band starts playing.
The Play button is then replaced by a transport bar, so there is only ever one control on screen that owns playback:
- ❚❚ Pause / ▶ Resume. A real pause, not a mute — nothing is synthesized while you are stopped, and the playhead is exactly where you left it when you resume.
- |◀ Back to start jumps to the top of the loop.
- ■ Stop ends playback and gives you the Play button back.
- The position slider and clock read 0:07 / 0:30 — where you are inside the loop, and how long the loop is. Drag the slider to move the playhead. Because live playback is a loop rather than a file, the clock wraps back to 0:00 and keeps going rather than stopping at the end.
The clock is not an estimate made on the page: it is the position of the audio actually being mixed, reported back from the audio thread about 29 times a second. It therefore reads what you are hearing, not what the engine is busy building a few hundred milliseconds ahead. Seeking is close to instant, and the short gap after you release the slider is the desk throwing away the audio it had already prepared for where you just were — better than a stutter of the old position.
While it runs, a line above the meters reads out what you are hearing — for example LIVE — C aeolian · 124 BPM · 4/4. When you change something musical it changes to LIVE — moving to …, at the next bar until the change actually arrives.
This is the same engine that makes the file, not an impression of it. Hold the board still and the stream matches the downloaded WAV to within about one bit of 24-bit audio — a difference some 130 dB below the music, which nothing can hear. It follows that mixing live and then pressing Generate gives you the file you were just listening to.
- Pressing Generate does not interrupt playback. The file renders in the background; when it finishes, the live desk quietly re-levels to match that file’s loudness.
- The stream loops your chosen length — which is what the transport slider spans. On the free tier that is 30 seconds. Change Length while it plays and the slider re-spans to the new loop at the next bar line, along with the music. It loops that one tune forever unless keep playing is ticked.
- Changing the seed is a different tune, but it does not stop the band: rolling the dice, or typing a seed and pressing Enter, recomposes at the next bar line like any other musical change, and re-renders the file if you had already generated one.
- The Loudness target is baked into the file, so it is not a live control: choose it, press Generate, and when the render lands the live desk jumps to the new level.
Keep playing: a different tune when this one ends
Tick keep playing, in the Playlist panel in the right-hand rail, and the desk stops being a loop and becomes a session. When the current tune has had its turn, a new one takes over at the loop point — a new seed on the same board, so your style, your faders, your key and every instrument you pinned all carry across. Only the tune is new.
- How long each tune gets. A tune repeats until it has held the room for about two minutes, then hands over: a 15-second loop plays eight times, 30 seconds four times, a minute twice, and a two- or five-minute track changes after a single play. The advance interval is not a hidden setting — it is your Length, rounded up to something you can work to.
- The board always names what is playing. The seed field and Copy link change at the moment the new tune is heard, not before, so a link you copy is always the tune in your ears. The line under the toggle tells you which tune is next.
- Heard something good? The Session panel keeps the last eight tunes of this session — one click puts one back on the board, and doing that turns keep playing off, because you asked to stay on that one. It is a shelf, not a library: to keep a tune past today, copy its link or save it as a preset.
- It never renders anything. Keep playing is listening state, like Volume and Mute — it is not in your links, your presets, your A/B slots or any certificate, and it will not start a render behind your back. If you type in the seed field it stands down until you leave, so it cannot overwrite what you are typing.
Shuffle: hand a control to the seed
Press Shuffle… in Setup and every control that can be re-rolled grows a small rnd pin. Mark the ones you want the seed to decide, press Done, and from then on those controls are re-rolled every time you roll the dice and every time keep playing starts a new tune. Everything you did not mark keeps exactly the value you set — that is the whole point of marking rather than of a "randomize everything" button.
- It is still deterministic. Nothing here is random in the sense that matters: every shuffled value is derived from the new seed, so a shuffled board is reproducible, and its link recreates it exactly, on any device, forever. The seed is the only arbitrary thing in the product, and it always was.
-
Your shuffle setup travels too. Copy link carries the marks alongside the board (as
&rnd=…), so you can hand someone the board and the parts you left to chance. It is not part of a preset or an A/B slot — those are boards, and this is a policy. - What cannot be shuffled, and why. Tempo is left out on purpose: pinning a tempo turns off auto tempo, which is what makes a track loop seamlessly, so a shuffled tempo would quietly break the loop. Length and Loudness are paid controls, and Length also owns the transport. Mutes are yours. A role's level is never re-rolled to zero, because zero is a mute rather than a quiet.
- Shuffling Style re-rolls the band. A new style brings its own instruments, scales, forms, chords and levels, so pins that belonged to the old style are dropped — exactly as they are when you change style by hand.
Playlists: a list of boards, played through
Save to my playlist — next to Copy link, in Keep this in the right-hand rail — puts the board you are hearing into a named list. The Playlist panel is directly underneath it, and holds keep playing and Shuffle… as well: everything that decides what plays next is in one place. You can also paste any Studio link into the panel, or press Fill with one per style to drop in the ten starting points at once. Each entry is stored as the same link a Copy link gives you, so a playlist track is a whole board: style, faders, key, instruments and seed.
- Playing one. Set Play to in order or shuffled and keep playing works through the list instead of drawing new seeds; the track you are hearing is highlighted. With Play set to off the list is just a shelf — press a track to hear it, or Next to step through by hand.
- Reordering and removing. The arrows move a track one place; ✕ removes it. Tracks are addressed by an identity of their own, so renaming or reordering never loses your place.
- It stays on this device. Playlists live in this browser's storage — there is no account and nothing is uploaded, which is the same trade every other saved thing in Studio makes. If the browser refuses to store one (a full disk quota), the panel says so rather than losing it quietly. To move a playlist somewhere else, copy the links out of it.
- Ambient Loops links are refused. The two boards' links look similar and share some keys, so an Ambient link would quietly parse into a wrong-looking Studio board. The panel rejects it instead.
What lands instantly, and what waits for the bar
Which controls are instant is a fact about the signal, not a design choice. Anything that is just a level on a finished part can be applied to the next few milliseconds of audio. Anything that changes which notes are played cannot — a note already sounding cannot be un-played — so it is cut in at the next bar line.
| When it lands | Controls |
|---|---|
| Instantly — the next audio block, about 3 ms, glided over roughly 10 ms so nothing clicks | Every role’s Level fader · Mute · Space · Stereo width · the desk’s level match after a render finishes |
| About four tenths of a second later | Brightness. It lives inside the sound-making itself, so it arrives with the next block of audio the engine builds — and the desk keeps a short queue of blocks ready ahead of the playhead, which is where the delay comes from |
| At the next bar line, crossfaded over a quarter of a second, continuing from the same point in the track | Energy · Motion · Swing · Key · Scale · Form · Chords · Tempo and auto tempo · Length · seamless loop · any strip’s instrument menu · Style (its notes; the new style’s levels and room land instantly) |
| When the tune ends | Keep playing's next tune — the only change that waits for the loop point rather than a bar line |
| Never in the file | Volume · Mute · keep playing. Listening state: it changes what you hear from this desk and nothing about what a render, a link or a certificate contains |
Because a change is cut in at a bar line and crossfaded, moving the key or the tempo mid-flight reads as a modulation rather than a glitch. It also means a musical change you make just after a downbeat waits nearly a whole bar — that is the control working, not lag.
The meters
Once live playback starts, a row of vertical bars appears under Setup — one per role in the current style, labeled with the role's name. They update about 29 times a second and fall away smoothly rather than flickering. They are rough loudness bars, not calibrated to any scale.
The meters read each part before its fader. They show what the band is playing, not what you are hearing. Pull a Level to zero, or mute the role, and its bar keeps moving. That is deliberate — it keeps the meters useful for finding which role is producing a sound you are trying to identify — but it does mean they are not a mix check. Trust your ears for balance.
A muted role’s meter is dimmed. The bar still moves, because the part is still being played; the dimming is there so a bouncing meter and a silent instrument never look like a fault. Dim and moving means “playing, but not in the mix”.
A bar that goes quiet for a while is usually the arrangement, not a fault. The engine thins the ensemble section by section: a lead sits out an intro, a pad drops in a bridge. No role is ever dropped from more than about 40% of the bars, and drums are never dropped at all — but how often a role actually plays inside the bars it is kept in is up to the part. Pads and leads rest. Measured across four seeds, Classical’s texture, Classical’s and Metal’s leads and the EDM and House rhythm parts all sat between roughly 55% and 65%. If a bar is dead for the entire loop, check whether that strip is muted or its instrument menu is on something very quiet.
Drums, Weather and Texture are the exceptions — they are never thinned and never rest. Texture is the room the piece is played in rather than a part played in it, so it carries straight through every section: rain does not stop between phrases. It also stays centred: a texture never slides across the stereo field. Any width you hear from it is the instrument's own, or Space. The one thing that can still leave a hole is a struck texture — Marimba rings and then decays, and at Ambient's slow tempos the notes can be five seconds apart, so you will hear the gaps between them. That is the instrument, not the arrangement; pick a sustaining texture if you want an unbroken bed.
A background layer also holds its level. Rain in particular does not swell and fade on its own: how loud it is answers to its own fader and the master volume, and to nothing else. If you hear the rain change, it is a fader moving — nothing in the instrument breathes.
Reading the plan
The plan is what the seed resolved to, inside the style’s rules. It used to be a readout in the
right-hand column; it is now carried in the certificate, which records exactly
the same fields. Open the downloaded .json in any text editor, or paste it into
Verify, and the plan is the plan object inside it.
While the desk is playing, the LIVE line under the transport gives you the two fields you most often want mid-session — key and tempo — without a render. The rest of the plan needs the certificate.
- Style — which pack played.
- Key — the tonic and scale the seed landed on, for example G mixolydian.
- Tempo — beats per minute and the time signature. Every style ships in 4/4 today, so this always reads 4/4.
- Bars — how many bars the track is. Multiply by the bar length to check the fit yourself: 12 bars at 96 BPM is exactly 30 seconds.
- Form — which template was used.
- Sections — the section list with its bar counts, for example intro:1 verse:3 chorus:3 …. This is where you see whether a song form actually had room.
- Chords — the first eight bars as roman numerals (see below).
- Band — which roles played.
How to read the chord row
Three different spellings turn up there, and they are all correct.
- Triad styles — Pop, Indie-pop, EDM, House, Ambient, Classical — follow the scale, so a minor track honestly reads i iv v i rather than claiming major chords, and a lydian Ambient track can read I vi iv° V I iv°.
- Rock, Punk and Metal write power chords, which have no third to be major or minor. Their rows print an upper-case numeral plus a 5 whatever the key — the board’s own default Rock render reads I5 - bVII5 - IV5 - I5, and a Metal track in F harmonic minor reads I5 - VI5 - VII5 - I5, flat-free, because those degrees are already flat in a minor scale even though the Chords menu names that same option i-bVI-bVII-i.
- Lo-fi’s seventh chords are spelled without the 7. Its ii–V–i vamp reads II V I I, and its suspension option reads I Isus4 IV IV.
Use the plan as a teaching tool: dice ten Rock tracks, read ten plans, and you will quickly learn which keys, tempos and forms you actually want to pin. Certificates are a licensed download, so this particular loop is a Studio-license one; on the free tier the LIVE line’s key and tempo are what you get.
The track editor
Studio generates an editable composition, not an opaque recording. Press Edit track… at the top of “Your track” and the editor opens on its own page, built for people who do not read music: no note names, no numbers, no decibels. You pick a stretch of the track, and then you change what the instruments play in it.
The whole track. The top of the page is a map — one row per instrument, one block per part of the track (Intro, Chorus, Quiet bit, and so on, named from the music the engine actually wrote). Taller marks mean an instrument is playing more. Click anywhere on a row to work on the part you clicked, drag sideways across a row to grab a longer stretch, or drag either ↔ handle. A green line runs across this map whenever anything is playing — the whole track or one section — so you always know where you are in the track. Scroll down and the map stays put at the top of the window, with the notes running behind it.
The notes. Every instrument in the track gets a row. Time runs left to right; inside a row, higher up is higher-sounding. Each instrument gets as many rows as its music needs, up to twelve steps of its own scale — and if it reaches further still, pushing a note off the top or the bottom scrolls the rows along to follow it. Drag a note to move it, or grab either end to make it longer or shorter. Pick a note and buttons appear: Higher · Lower · Louder · Softer · Longer · Shorter. It always lands on a note that fits the rest of the track, so it cannot sound wrong. Drums work the same way, except their rows are named — Kick, Floor tom, Snare, Hi-hat, Crash. Beside each row is that instrument's own column: a box, its name, and its instrument menu. Click the name to point the controls on the right at it. Untick the box to leave it out while you listen — it drops out immediately, without stopping the music, and it is listening only: it changes nothing about the track, nothing in the save, and nothing the board renders or downloads. The menu swaps the instrument and keeps every note; the option with an asterisk is what the board was playing, so you can always find your way back.
Sliding between two notes. Click a note, then hold Shift and click any later note of the same instrument — or press Shift + Enter on it, if you are working by keyboard: the first one slides into that note's pitch instead of holding its own. It does not have to be the very next note; on an instrument playing chords, where several notes start at the same moment, it usually should not be. A dashed line appears across the sliding note showing the path the pitch takes — drag the dot on it, or use the up and down arrows, to shape the slide from "moves straight away and holds" to "waits, then swoops at the end". Repeat the gesture to take the link away, or shift-click a different note to re-aim it. Move the note it points at and the slide follows it; delete that note and the slide goes with it.
Volume. In the panel on the right, a line you bend rather than a number. Drag any of the five dots, or take a ready-made shape: Level, Fade in, Fade out, Swell, Dip, Build. It applies to the instrument the panel is following, across the whole of the part the selection sits in — left is the start of that part, right is its end — and it lands at the next bar, like every other change to the notes. Across a whole track that is a subtle change; within the part it governs it spans roughly −9 dB to +5 dB.
How it's played. The controls on the right are named for what your body would do — a strummed instrument gets how hard it is strummed and how long notes ring; a drum kit gets how hard it is hit and timing, loose to tight; a bowed instrument gets bow pressure. The panel names the family and the instrument it is following — drums · Acoustic kit — so you always know which of your instruments the controls are shaping. Some layers — rain, thunder, vinyl crackle — are the room itself rather than a part someone plays, and the editor says so instead of offering controls that would mean nothing.
It keeps playing while you work. The editor has the same live desk the board does: press Play and the band starts, and it does not stop again while you change things. Mute, and the room and width the board was set to, land instantly. Anything that changes the notes — moving one, swapping an instrument, bending the loudness line, linking two notes — is cut in at the next bar line, which is where a musician would put a change anyway. Nothing re-renders and nothing waits.
Nothing is permanent. Play plays the whole track. Play section — in the row directly above the notes, with Undo and Put this part back, and the times and bar count of what you have selected — plays only those bars; it stays put at the top of the screen as you scroll through the notes. There is one output, so starting either one stops the other, and Loop beside them repeats whatever is playing until you stop it. Volume and Mute, beside Play, are the same listening controls the board has — they change nothing about the track. Undo steps back one change at a time. Put this part back returns every instrument in the selected bars to exactly what the engine wrote — your edits elsewhere stay. Everything is saved on this device, so you can close the page and come back.
Getting the audio. The editor has no download of its own: go back to the board and press WAV in its Download box, which renders your edited version. What you hear in the editor is a live stream of the same music, so the two agree musically — but the file is rendered in one pass and the stream is composed as it plays, so they are not identical to the last decimal place. The file is the master. An edited track cannot carry a certificate, and the board will tell you so rather than issue one — a certificate proves that re-rendering the seed and settings reproduces the audio, and that is true of the generated version, not of the one you changed. Put the part back and it can be certified again.
You can do all of it without a mouse: Tab reaches each instrument's box, name and menu, both selection handles, every note and every dot; arrow keys move whatever is focused; Shift with left and right makes a note shorter or longer. Every change is read out as you make it.
The grid itself — tempo, meter, how many bars — is fixed when the track is generated, and the key, seed and style stay on the board. The editor shows them as context and never duplicates them.
What Studio hands you
Your track is always on the board, dimmed until you render something. Press Generate track at the top of it and its Download box lights up. Everything Studio can give you lives in that box — there is no file player here, because the desk already plays this board live.
| Output | What the file actually is | Cost |
|---|---|---|
| WAV | The finished stereo track — 24-bit, 44.1 kHz, uncompressed | Free up to 30 seconds |
| Certificate | A small text file describing exactly how that audio was made, with fingerprints that let anyone re-create it | Studio license |
| MIDI | The notes, with no sound — one track per role, for your DAW | Studio license |
| Stems | One WAV per role plus a reverb-only track, to mix yourself | Studio license |
| Loudness | A menu that bakes a loudness target into the WAV before you download it | Studio license |
One rule runs through all of it: the WAV, the certificate, the MIDI and the stems all describe the track currently on screen. If you move a fader or change the key after generating, the exports still match what you can hear until you press Generate track again.
WAV
The plain, finished file: 44.1 kHz, 24-bit, stereo, uncompressed — the same format at every length and on both tiers. There is no watermark, no downgrade, and no difference in quality between a free download and a licensed one. Only the available lengths differ.
It is named after the board that made it — seed, then style, then length, for example amber-drift-204-rock-30s. That name alone is enough to find the track again, because the seed plus the style plus the settings is the track.
| Length | File size |
|---|---|
| 15s | 3.8 MB |
| 30s | 7.6 MB |
| 60s | 15.1 MB |
| 2 min | 30.3 MB |
| 5 min | 75.7 MB |
That is roughly 15 MB per minute. If you need an MP3 or AAC, convert it after downloading — Studio only writes WAV.
Rendering takes real time, and the status line counts up a percentage while a Cancel button sits next to Generate. A five-minute track is about ten times the work of a 30-second one, and a phone or an older laptop will be slower than the numbers quoted earlier in this manual.
Certificate
The certificate is a small text file — around 3 KB — that you download next to the WAV, named after the same board. It records:
- the engine and its exact version, and your seed;
- every setting that shaped the render: style, length, sample rate, bit depth, seamless loop, all six Feel sliders, every role’s level and reverb send, plus any key, scale, tempo, form, chord progression, instrument pin or loudness target you chose;
- the resolved plan — the key and scale it settled on, tempo and meter, number of bars, the form and its sections, the chord progression in roman numerals, which instrument played which role, and which drum patterns were used;
- the audio’s own facts: sample rate, channels, length in frames and seconds, bit depth, and two cryptographic fingerprints — one of the audio samples, one of the whole file;
- a written statement of how the audio was synthesized — deterministically, from the seed and settings recorded above.
The LIVE line on the stage is a two-field excerpt of that same certificate — key and tempo — so a free user can see what the seed chose there, and a licensed one can download the file that documents all of it.
What a musician does with it. Keep it in the same folder as the WAV. Hand it to a client, a label, a stock platform or a rights-management team as documentation of where the audio came from; anyone can check it themselves at Verify, with no account and nothing uploaded.
Two honest limits. The certificate is not signed by anyone and carries no timestamp. It does not prove when you made the track, or that you were the one who made it. Its entire force is that the audio can be regenerated from it, bit for bit, by anyone — a statement about origin, not about ownership or date.
MIDI
MIDI export gives you the notes, not the sound: one standard MIDI file at 480 ticks per quarter note, named after the seed and style, a few kilobytes in size. Pressing the button downloads it immediately.
Inside:
- a first track carrying the tempo and the time signature, so your DAW’s grid lines up straight away;
- then one track per role that actually plays, named after the role and the instrument the seed chose;
- anything struck — the kit, and shakers or tambourines wherever they sit — on the standard General MIDI drum channel and key map: kick 36, snare 38, closed hat 42, open hat 46, crash 49, ride 51 and so on, so any drum plugin reads them correctly. Routing follows what the note is, not which strip it came from, which is why Weather does not land there: a thunder roll is a held pitched note, so it gets a melodic channel of its own like any other sustained part;
- pitched roles each on their own channel;
- the composer’s own note-by-note dynamics, mapped across the full velocity range.
What it does not carry. No sound, no synth settings, and no instrument selection: the file contains no program changes, so your DAW opens every pitched track on whatever its default instrument is. That is the point — the MIDI is a sketch to re-voice with your own instruments, while the certified WAV stays exactly what it is. Levels and mutes do not reach MIDI either; a muted role is still exported in full, with its dynamics untouched, because the mix belongs to the mix and MIDI carries the score.
One detail worth knowing: where a role holds the same pitch across a bar line, the overlapping repeats are merged into a single sustained note. A pad that holds a chord over four bars arrives as four long notes, not sixteen stuttering ones.
MIDI is a derived file. It is not covered by the certificate, which describes the audio.
Stems
Stems give you the track as a multitrack. Press Stems and, after a short render, a row of download buttons appears under the Get box — one per playing role, plus one called reverb. Click each one to save it.
Which stems you get depends on the style, because each style plays a different band.
| Style | Role stems (plus the reverb return in every case) |
|---|---|
| Pop, Indie-pop, Rock, EDM | drums, bass, rhythm, harmony, lead |
| Punk, Metal | drums, bass, rhythm, lead |
| House | drums, bass, rhythm, harmony |
| Lo-fi | drums, weather, bass, rhythm, harmony, lead, texture |
| Ambient | weather, bass, harmony, lead, texture |
| Classical | bass, harmony, lead, texture |
A role whose level is at zero — fader all the way down, or its strip switched off — produces no stem file at all.
Three things to know before you open them in a DAW.
- Each role’s fader position is baked in. The stems print your mix, not a flat one. If you want to rebalance from scratch, set the role levels where you want them (or all equal), press Generate track, then export stems.
- The role stems are dry. All the reverb in the track is in the single reverb file. Bring it up under the dry stems to restore the space you heard, or leave it out and use your own.
- They sum to the mix before the final master stage. Summing every role stem plus the reverb at unity gives you the track as it stood before stereo width, the safety clipper, the peak ceiling and any loudness target were applied. That is very close to the downloaded WAV but deliberately not identical — it is a little hotter and rawer. Treat stems as a multitrack to mix, not as a way to take the finished file apart.
Watch the size. Each stem is a full-length 24-bit WAV, the same size as the mix — six stems of a five-minute track is roughly 450 MB. Stems at 30 or 60 seconds are a much friendlier way to work. Like MIDI, stems are not certified; the certificate describes the mixed WAV.
Loudness — when to use a target instead of Natural
The Loudness menu sits in the Get box and offers four settings.
| Setting | Use it for |
|---|---|
| Natural (default) | Anything you are going to mix, master, or hand to an engineer |
| −14 LUFS | Uploading straight to YouTube, Spotify and similar |
| −16 LUFS | Podcast delivery |
| −23 LUFS | Broadcast delivery (EBU R128) |
Loudness is baked into the file, so changing the menu does nothing until you press Generate track again — the status line says so when you change it.
Natural is a true bypass: no measurement, no gain, nothing touched. It is not “quiet”, either. Natural renders land roughly in the streaming zone already — measured on short test renders across two seeds each, rock, metal, EDM, ambient and classical all came out somewhere between about −13 and −17 LUFS, and the exact figure moves by a decibel or so with the seed. For most work Natural is the right answer, and the targets are for delivery.
The rule that matters: any target quieter than the render’s natural loudness is hit exactly; any target louder than it can fall short, because the loudness stage is never allowed to push the file into clipping. In testing, only −23 landed exactly in every style. Pop rendering naturally at −17 to −18 landed at −16.5 and −17.4 when asked for −16; house −17.1 landed at −16.2; EDM −16.7 landed at −16.0 and −16.3. The file is never louder than you asked and never clipped; it can be a fraction quieter.
Lo-fi is the case to watch. It renders naturally at about −22.6 LUFS, six to eight units under rock, so asking for −16 lands only around −19 to −20, and asking for −14 gives you the same file. If you need lo-fi at podcast loudness, raise Energy, pull nothing down, and normalize in your own tools where you can also control the peaks.
Whichever target you use, the certificate records both the target you asked for and the exact gain that was applied — so the loudness move is part of the provable history of the file, not something done to it afterwards. On the free tier the menu stays on Natural: choosing a target opens the license panel and the menu snaps back.
Session: A/B slots and presets
The A and B slots
The Session panel keeps two whole-board snapshots for comparing arrangements.
- Set A / Set B capture the board as it stands. Both need the Studio license.
- Apply A / Apply B put a captured board back, and stay greyed out until that slot has something in it. These need the license too.
A slot stores the board, not the audio: seed, style, length, seamless loop, all six Feel sliders, whatever you pinned for key, scale, form, chords and tempo, the loudness target, every role’s Level (with muted roles stored as zero), and every instrument pin. It does not store the rendered track or its certificate — applying a slot never renders by itself, so press Generate when you like what you see.
Both slots live in your browser on this device. They survive a reload, are not synced anywhere, and are gone if you clear site data. A slot captured with the license and recalled without it comes back clamped to 30 seconds.
The ten starters
Below the A/B row is a list of preset buttons. The first ten are starters — one per style, all at 30 seconds. Loading any of them is free, they cannot be deleted, and they are a good way to hear what each style does before you touch anything.
| Starter | Style | Starter | Style |
|---|---|---|---|
| Nebula | Ambient | Neon Drive | EDM |
| Daydream | Pop | Midnight Loft | House |
| Golden Hour | Indie-pop | Chamber | Classical |
| Arena | Rock | Study Hall | Lo-fi |
| Basement | Punk | Forge | Metal |
A starter sets its style, seed, 30-second length, seamless loop and Energy. Loading one also brings back that style’s own defaults: its swing, its preferred tempo with auto tempo back on, Auto for key, scale, form and chords, and every role strip at the style’s default Level with its menu on Auto. Loudness returns to Natural. Motion, Space, Brightness and Stereo width stay exactly where you left them — a starter is a starting point, not a factory reset.
Your own presets
Saving a preset needs the Studio license — but loading and deleting presets does not. Type a name and press Save preset; leave the name blank and it is numbered for you. Saving under a name you have already used replaces it. Your presets appear after the starters, each with a ✕ to delete it. They store precisely what an A/B slot stores — the board, not the audio — and live in this browser on this device. Loading a preset never renders on its own.
If you want something you can send to somebody else, or keep outside this browser, use Copy link at the top of the console instead.
Free versus licensed
Studio’s whole instrument is free. The locks are almost entirely on outputs.
Free, with no code and no account:
- All twelve styles
- Every control on the Feel, Musical and Roles panels — Energy, Motion, Swing, Space, Brightness, Stereo width; key, scale, form, chords, tempo; every role’s level, mute and instrument menu
- Style, seed, seamless loop, and lengths of 15 and 30 seconds
- Play live — the live mixing desk with its per-role meters
- Unlimited generating, and a full-quality 30-second WAV
- The LIVE line — the key and tempo the seed chose, while it plays
- Copy link — share a whole board with anyone
- The ten starter presets, and loading or deleting any saved preset
Studio license — $49, one time:
| Locked behind the license | What it means in practice |
|---|---|
| Lengths 60s, 2 min, 5 min | Free tracks cap at 30 seconds |
| Certificate download | The generation record next to the WAV |
| MIDI export | The score, for your DAW |
| Stems export | The multitrack, plus the reverb return |
| Loudness targets | −14 / −16 / −23; free stays on Natural |
| A/B compare | Both capturing a slot and recalling one |
| Saving your own presets | Loading the starters, and loading or deleting saved presets, stays free |
| Commercial use | A commercial-use license covering the audio in paid work |
Locked controls wear a small Pro badge and stay clickable — clicking one opens the license panel rather than doing nothing.
The length and loudness limits are real, not just greyed-out buttons. They are applied at the moment of rendering, so a shared link, a saved preset or an A/B slot that specifies five minutes and −14 LUFS renders 30 seconds at Natural for a free user — and the board honestly shows 30s and Natural rather than pretending otherwise.
Price: Studio is $49, one time, never a subscription, and it also covers the Ambient board, so it is the single license for the whole site. The Ambient-only license is $19, one time. If the store has not been switched on yet, the license panel says commercial licenses open soon and that the tool stays free in the meantime — that message is about the license; the Pro outputs stay locked until a code is entered.
Unlock codes, and why entitlement is per device
There is no account and no login. Buying gets you a code, and the code is the license.
- A Studio code looks like STU-XXXX-XXXX; an Ambient-only code starts with AMB-. The characters avoid O, 0, I and 1 so nothing is ambiguous when you read it out of an email. Capitals and lower case both work, and spaces are ignored.
- To use it: open the license panel — the Studio license — $49 button in the Get box, or any locked control — expand Already have a code?, type the code and press Unlock. Everything unlocks on the spot with no reload, and if a track is already on screen its Certificate download lights up immediately.
- The example shown inside the code box reads QT-XXXX-XXXX. That is an old placeholder; your Studio code begins with STU-.
- An Ambient-only code entered on Studio politely tells you it unlocks the other product, and is not stored — it cannot overwrite a Studio code. A Studio code entered on the Ambient board works there.
Entitlement is per browser, on one device. The code is remembered in that browser’s storage for this site and re-checked every time the page loads. That means:
- A different browser, a different computer, or a phone will each need the code entered again — so keep the code, not just the receipt.
- Clearing site data removes it, and a private window forgets it when closed.
- There is no device counter. Comp codes are checked offline, on your machine. Redeeming or recovering a code from a Stripe purchase contacts the license server with the code or email you enter — and nothing else.
Certificates: what one proves, and how to verify it
Studio’s engine is deterministic: the same version, the same seed and the same settings always produce identical audio, sample for sample. A certificate records exactly those inputs plus a fingerprint of the resulting audio, so anyone can regenerate the track from the certificate alone and check that the fingerprints match. Audio taken from anywhere else cannot survive that test, because regenerating from the certificate produces something different.
It proves origin — this audio is the output of this engine from these inputs. It does not prove when you made it and it does not prove who made it.
How to verify one
Go to Verify, which is linked from the Get box and the site nav.
- Click the Certificate box and choose the certificate file. The page shows the engine version and seed it read.
- Optionally click the Audio box and add the WAV. This is worth doing — it is what proves the file in your hands is the certified one, rather than merely that a certified file exists.
- Press Verify. The page re-renders the entire track in your browser, which takes about as long as generating it did, then compares fingerprints.
Nothing is uploaded. The certificate and the audio are read locally, the engine runs locally, and the site has no server to receive them.
The three checks it reports
- Re-render — the audio was regenerated from the certificate’s inputs and compared. The core check.
- File match — the WAV you supplied was fingerprinted against the certified fingerprint. Shown as a dash, meaning skipped, if you did not supply one.
- Version — whether this build contains the engine version that minted the certificate.
| Verdict | What it means | What to do |
|---|---|---|
| ✓ Valid | The audio was regenerated and matched. If you supplied the WAV, it is the certified file. | Nothing — this is the good outcome |
| △ Unverifiable here | This build does not contain the engine version that minted the certificate, so the audio was not examined at all. | Verify with a build that includes that version. This is not a forgery finding |
| ✕ Invalid | Either the re-render did not reproduce the certified audio — the certificate has been altered, or the audio is not the output of those settings — or the WAV you supplied is not the certified one. The line under the table says which. | Check you paired the right WAV with the right certificate before drawing conclusions |
About “Unverifiable here” — read this before you panic. A certificate names the engine version that made it, and verification re-renders with that version, not the newest one. Whenever the engine changes in a way that alters the audio, the version number is deliberately bumped, precisely so that an old certificate reports “unverifiable here” instead of a false mismatch that would brand authentic audio as forged. The site’s build today carries Studio’s current engine, version 2.4.0. A certificate minted by an earlier Studio build will therefore read △ Unverifiable here: the audio has not been judged, the verifier simply cannot run that version. The intent is that every released version stays available so certificates remain checkable indefinitely.
If a platform flags your track, give them the WAV and its certificate and point them at the Verify page. They can run the check themselves without installing anything or sending anyone a file.
The license in plain language
Free use. Both tools are free — every style, the full board, the live desk, and a 30-second WAV to download. Free audio may be used in personal, non-commercial projects: hobby videos, personal streams, study playlists, prototypes, class projects. No attribution required.
Sonodrift Studio — $49, one time. Unlocks the full Studio and grants a perpetual, worldwide right to use your generated audio in commercial projects, with no limit on how many projects or how much revenue: client work, monetized video and podcasts, livestreams and VODs, games, apps, in-store or venue playback, film and advertising. It also includes everything an Ambience license covers.
Sonodrift Ambience — $19, one time. Commercial-use rights for the Ambient board only. That board is free to use either way; this buys a commercial-use license for its output, not extra features.
Both are one-time purchases, never subscriptions — deliberately, so work you have already published can never retroactively lose its license.
Royalty-free means the license charges no per-use royalty: pay once and use your exports in your projects without counting uses, views, or revenue. It does not mean copyright cannot exist in a track, or that no platform or third party can ever make a mistaken claim.
The one thing you may not do, free or licensed: resell or redistribute the raw audio files as audio — republishing generated tracks as a stock-music library, a sample pack, or a competing catalog. Use them in things; do not sell them as the thing.
Sonodrift outputs are nonexclusive: the same or a similar track may be generated by or licensed to someone else. Do not enroll an unmodified export — or its stems, loops, or lightly modified versions — as a reference in Content ID, Rights Manager, or any other automated copyright-claiming system, and do not represent it as exclusively yours. Your completed project remains yours to register or claim to the extent the law permits; exclude the Sonodrift source audio from any automated exclusive reference.
How the audio is made. Every track is synthesized in your browser, deterministically, from your seed and settings. The certificate documents how it was produced, and re-rendering reproduces the identical file.
The honest fine print, as written on the License page: it is not legal advice and is no substitute for it in high-stakes uses; if a platform wrongly flags your track, your certificate documents its provenance to help you contest the claim; certificates are meant to stay verifiable indefinitely, because every released engine version is kept available for re-rendering; and if the license does not do what the page says, tell them within 14 days for a refund.
Recipes
Style Lo-fi, Key C, Scale Minor (aeolian), Chords lofi-ii-V-i, Length 30s, seamless loop on, auto tempo on. Press Play live. The LIVE line will read C aeolian at 72, 80 or 88 BPM depending on the seed, and the chords repeat ii–V–i for the whole track. Listen for the vinyl crackle under the electric piano and the heavy drag on the offbeat sixteenths. Dice a few times: the key, the chords and the loop stay put; the tempo, the form and the instruments move.
Style House, Key A, Scale Dorian, Chords i-IV, Length 30s, seamless loop on, auto tempo on. Generate. Every seed lands on 120 or 128 BPM because the range is narrow, the piano stabs alternate two chords, and the loop point falls exactly on a bar line. Copy the link before you move on — that is your saved version.
Type a seed you like, leave every Musical menu on Auto, and press Generate. Then change only the Style and press Generate again, ten times. The key stays the same in all ten, because the seed decides it and the style does not, while the tempo, the scale, the form and the whole ensemble change underneath it. It is the fastest demonstration of what the board actually is.
Pick a style, leave seamless loop checked and auto tempo on, choose 30s, Generate, and press WAV. Drop it on your timeline and repeat the clip; the joins will be inaudible. Free downloads are licensed for personal, non-commercial use.
Generate a rock track you like, press MIDI, and drag the file into your DAW. You will see drums, bass, rhythm, harmony and lead as separate tracks at the right tempo. Mute everything except the lead, point it at your own synth, and play it against the downloaded WAV to keep the rest of the arrangement.
Choose a style, set the drums fader where you like it, Generate, then Stems. Download the drums stem and the reverb stem, drop both into your session, and you have the kit dry with its room available separately.
Limits — things Studio does not do
- Every style is in 4/4. No 3/4, no 6/8, no odd meters ship today.
- Rendering takes real seconds, not milliseconds. There is a percentage read-out and a Cancel button for a reason.
- Only a seed change re-renders by itself. Move a slider, change a scale, switch a style — the board updates but the file on screen does not until you press Generate track. During live playback the changes are audible immediately or at the next bar, but the downloadable file still needs a Generate.
- No solo button. Mute the rest; it is instant while live.
- No per-role pan, EQ or compression. A strip is Level, mute and instrument. Pan is fixed by the style, and there is no EQ in this engine at all. For real mixing, export the stems and mix them where you normally would.
- Reverb sends are automatic. Each role’s send is the style’s and is not exposed; Space scales the whole room.
- Percussion appears as Weather on Ambient; Counter is still unused. The engine has always reserved both slots. Ambient now fields the first as its thunder layer — a style may rename a slot when the generic name would mislead — and no shipping style uses Counter, so you will not see that strip.
- The instrument menu cannot borrow across styles. Each role offers what its style allows; a pin from another style is quietly ignored and that strip falls back to Auto.
- Meters are pre-fader, so they do not reflect your mix.
- Studio only writes WAV. Convert to MP3 or AAC afterwards.
- Some menu entries show raw names. Lo-fi’s Form and Chords entries keep a “lofi-” prefix, and Ambient’s progressions are numbered 0 to 7. Cosmetic, but it is what you will see.
- The live preview needs a capable browser, and can stall on a slow device. The downloaded file never does.
Troubleshooting and FAQ
I moved a control and the track did not change.
Almost always: you have not pressed Generate track since. Only a seed change re-renders on its own. If you have re-generated, check whether the control is one of the known-inert cases: Swing does literally nothing on Ambient and Classical, Motion is close to inaudible on the band styles and on takes under a minute, and Energy drops no instruments on a seed that drew a loop form. The upper half of Energy also does nothing to EDM, House, Metal and Punk, whose peak sections already sit at the ceiling.
The track suddenly got much louder and started sounding harsh.
Brightness above roughly 0.6 on Rock, Metal, Punk or Classical. That is the clipper, not extra treble — bring it back to 0.5 to 0.6. See Brightness.
My loop clicks, or the beat stumbles at the join.
You have pinned a tempo that does not divide the length into whole bars. Tick auto tempo, or choose a length your tempo divides evenly. On Ambient and Classical there is no exact fit at all, but they are beatless so the join is rarely audible.
The same seed gave me a different track.
Seeds are case- and space-sensitive, so check the spelling exactly. Then check the rest of the board: a seed only fixes the decisions left on Auto, and the style, length, Feel sliders and pins are all part of the track. Across engine versions the audio can change deliberately, which is why certificates name the version that made them.
My longer length or loudness target reset itself.
You are on the free tier, and a link, preset or A/B slot asked for something the free tier cannot render. The board shows you 30s and Natural honestly rather than pretending. Enter a Studio code and it will hold.
My swing setting disappeared.
You changed style. Swing snaps to the new style’s authored value every time. Copy a link or save a preset before style-shopping; both restore the exact value.
A role strip vanished from the mixer.
Each style fields a different band. Punk and Metal have no Harmony, House has no Lead, and Ambient and Classical have no Drums. See Roles.
A meter is moving but I hear nothing from that role.
The meters read each part before its fader, so a muted role or one pulled to zero still shows movement. Check the strip’s checkbox and its Level.
“Audio is stalling on this device.”
That message is exact: a stall affects only the live preview. Your download is unaffected, because the file is computed offline at its own pace. Try a shorter length for live playback.
Play live will not start at all.
Live playback needs modern browser audio. On an old browser or a locked-down embedded view, Studio falls back silently to playing the rendered file, and the Play button says live mixing is unavailable. Every control still works; it just needs a Generate to be heard. If nothing has been rendered yet, the status line asks you to generate a track first.
My certificate says “Unverifiable here”.
That is not a forgery finding. It means this build does not contain the engine version that minted the certificate, so the audio was not examined. Verify it with a build that includes that version. See Verifying a certificate.
I entered my code on my laptop and my phone still says Pro.
Entitlement is stored per browser, per device, with no server involved. Enter the code again on each browser you use — and keep the code itself, not just the receipt.
Why don’t the stems add up exactly to the WAV?
They sum to the mix as it stood before stereo width, the safety clipper, the peak ceiling and any loudness target. Stems are a multitrack to mix, not a way to reverse-engineer the finished master.
I asked for −16 LUFS and got about −19.
That will be Lo-fi, which renders naturally around −22.6 LUFS. A target louder than the natural render can fall short, because the file is never pushed into clipping. Raise Energy, or render Natural and normalize in your own tools. See Loudness.
Can I get 3/4, an MP3, or a track longer than five minutes?
No to all three today. Every style is 4/4, Studio writes WAV only, and five minutes is the longest length.
Do I have to credit Sonodrift?
No. No attribution is required, free or licensed.
Does the seed mean anything musically?
No. It is just a label. There are no good or bad seeds — only ones you happen to like, which is why the dice is the fastest control on the board.