What you can ask for
Written the way you’d say it, not the way it’s implemented. Each entry is something Tine performs in your open set — and where a capability has a limit, it says so here rather than after you’ve bought.
30 of the 38 show tested phrasings: those exact words are run against a real Ableton session by our test suite, on every model below, and what Tine said back is quoted verbatim — transcripts, not examples.
They are a floor, not a syntax. You can be much vaguer than these, and how vague you can be scales with the model you run: a small local model wants a clear verb and a clear target, a larger one follows pronouns and implied targets, and the hosted model takes the most roundabout phrasing and the most compound requests. The capability underneath is identical — what changes is how much interpretation sits in front of it.
Playing & recording
Start and stop playback
One word each. "Play" and "stop" run the transport in about a third of a second, without reaching for the space bar — useful when both hands are on an instrument and the keyboard isn't.
tested phrasings- “play”
- “stop”
- “start playback”
- “hit play”
- “stop the transport”
The metronome, in one word
Say “metronome” and the click comes on; say it again and it’s off. A button, not a sentence — it answers in under half a second, and it reads the click’s real state first, so it still works after you’ve toggled it by hand.
tested phrasings- “metronome”
- “metronome off”
- “turn the click on”
Set the tempo
Change the project tempo by saying the number, mid-session, without stopping what you're playing.
tested phrasings- “set the tempo to 90”
- “change the tempo to 128”
- “make it 100 bpm”
Launch clips and scenes
Fire a named clip or a whole scene, and stop them again, without hunting for the right slot in a full Session grid.
tested phrasings- “fire the seed clip on the RecScratch track”
- “stop all clips”
Arm a track and set monitoring
Arm the track you're about to play and set its monitoring, so the take starts on your terms instead of after a detour to the mouse.
tested phrasings- “arm Lead”
- “arm the Lead track for recording”
- “record enable the Lead track”
Record a take over a bar range
Punch in and out over a range you name out loud — "bars 5 to 8" — with the transport, the punch points and the record mode all set for you. This is the operation voice wins outright: your hands are already committed.
tested phrasings- “record a take on the RecScratch track”
- “record a take over bars 5 to 8 on the RecScratch track”
- “start recording a new clip on the RecScratch track from bar 4”
Choose between your takes, hands free
limitsPlay a few passes at a part, then pick one out loud — "keep the last take" — without stopping to drag anything. You can audition first ("play take 1"), and the take stays the subject afterwards, so "keep that one" means the one you just heard. Nothing else is thrown away: the other passes stay in their lanes.
tested phrasings- “keep the last take”
- “keep the first take”
- “keep take 2”
- “play take 1”
Move between Session and Arrangement
Switch views and hand playback back to the timeline, so you actually hear the arrangement again after working in Session.
tested phrasings- “switch to arrangement view”
- “take me back to session view”
And the things it deliberately doesn’t do
A question is not an instruction, and thinking out loud in front of an open microphone is not a command. An assistant that edits your set because you wondered something aloud is one you switch off — so these are tested with the same weight as everything above, except the test asserts that nothing changed.
A question, not an instruction. Tine answers it and leaves your set alone.
Asking about a feature is not asking for it to be applied to the clip you happen to have selected.
Answers the question without switching your view to demonstrate.
Thinking out loud in front of an open microphone. Nothing is added, nothing is changed.