Recursive Creative Systems works inside Ableton Live. It listens to the room, reads up on the record, plans, produces what is missing, sets the tone, checks its own work, and keeps what you decided. In the studio and on stage, with you.
Nothing in that was guessed.
The tempo came from what it heard, not from a website. The tone came from measuring the guitar on the record you were playing, not from a preset with a band's name on it. And it did not say “done” until it looked at Live again and saw the take where it said it would be.
Other tools turn a sentence into a clip. This one turns your room, the record and your history into a session you can play in.
Every session teaches it something: the artists you return to, the scales and keys you reach for, the way you like an EQ, what you said no to. It does not start over when the window closes.
That memory is what turns a set of tools into a collaborator. The orchestrator plans with it, the instruments read it, the generators follow it. It is yours: it exports as one document and imports back anywhere.
The work you would rather not do.
Placing takes, building beds, matching a tone, setting a count-in, naming scenes, keeping the set tidy. Under your policy: ask first, do it, or never.
The work you want a second opinion on.
Eight instruments answer as experts when asked. A mastering chain, an EQ, a restorer, a voice tool, each with a measurement in hand and a proposal in Live's own names.
The parts that are missing.
Quick follows your own recording and completes a part in seconds. Deep composes a whole song, with singing when you want it. Both know the tempo and key because they were measured, not typed.
With you, live.
A count-in, the bed, the tone, then you play. The same system is being extended past the studio to the rooms a show is played in: lights, cues and timecode from the same conversation. That part is in the works.
Ask for a sound and it researches the record, writes a plan you can read, and asks to listen to the exact thing before touching the set. What it knows on paper and what it heard in your room are kept apart, and both are said.
After anything that changes the set, it observes Live again and compares. The clip landed. The chain changed. The track is gone. If not, it says so. Every turn is one undo step, so a whole tone comes back with one key.
Tempo, deleting, playback and automatic changes each need your say-so. It never fires a clip you did not ask for.
Each one is a trained model with a face in Live. Each one can describe what it hears and propose what to change. Control applies the proposal under one undo step, by your policy.
A fifteen-band EQ that matches a reference or suggests its own curve.
nowEQ into compressor into limiter, fitted to a reference or set by its own model.
nowDeclip and denoise in real time, and deeper repair offline.
nowNoise reduction trained on music, so it keeps the character of what it cleans.
nowPitch correction and voice conversion, each on its own, never forced together.
nowTempo-synced delay and reverb.
nowLoudness balancing across channels, into a look-ahead limiter.
nowA plug-in that asks for audio to exist: a part, a variation, a cover, a stem, from a brief and your references.
nowAnother maker's instrument, in the same room: amp models, and captures of your own amp.
nowEvery capture, generation, decision and turn is kept as an event in your account, with the set it happened in. A project is the set you have open; the next session on it starts where the last one ended.
No instrument holds your keys. They read your identity through Control, and Control asks you before it changes anything.
format rcs-identity/1 plays Telecaster, single coils setup Ableton Live 12, Scarlett 2i2 artist The xx studies since 2026-09-10 tone clean, small speaker prefers eq never past ±6 dB prefers project xx.als 3 decisions, 2 captures
There are several ways now to type a sentence and get a clip in Live. They stop at the same place: they cannot hear the room, they cannot check what they did, and they forget you when the window closes.
| Ability | Chat automation for Live | Recursive Creative Systems |
|---|---|---|
| Turns a sentence into notes or clips | yes | yes |
| Listens to what is playing and measures it | no | capture, tempo, key, chords, stems, tone |
| Reads up on the record before acting | no | with the plan shown |
| Checks its own result in the set | no | observes Live again after every change |
| Instruments that are trained models | no | eight, in the room |
| Learns how you decide | no | an identity that is yours |
| Open to other makers' instruments | rarely | a seat in the room |
Control is the base: the orchestrator in Live, the listener, the researcher, the checks. Add instruments when you want them. Sign in with your email for Quick, Deep and your identity.