Question 1

What is Tine?

Accepted Answer

Tine is a desktop assistant for Ableton Live. You say or type what you want — "warp that to transient and loop the first bar", "sidechain the bass to the kick with a 30ms release", "render every clip on this track to stems" — and Tine drives Live directly. It is not a plugin and not a chatbot that gives you instructions: it performs the operation in your open set.

Question 2

Does it write music for me?

Accepted Answer

No, and it never will. Tine is a tool, not a generator. It doesn't compose parts, invent melodies, or produce audio out of a model. Everything it does is an operation you could have done by hand — it just removes the clicking between your idea and your session. The music stays yours.

Question 3

What can it actually understand?

Accepted Answer

Everything Tine does is listed on the "what you can ask for" page, in plain language: transport and recording, clip and note editing, warping, device and parameter control, mixing and sends, arranging, library loading, plus native operations like headless save, render, and stem separation. Many entries show phrasings marked as tested, which means those exact words are run against a real Ableton session by our test suite before each release.

Question 4

Do I have to use the exact tested wording?

Accepted Answer

No. The tested phrasings are the ones we can prove, not the only ones that work — think of them as a floor rather than a syntax. In practice you can be considerably vaguer than those examples, and how vague you can get scales with the model you run. A small local model wants a clear target and a clear verb; a larger local model handles pronouns, implied targets and two requests in one sentence; the hosted model handles the most roundabout phrasing and the most compound requests. The capability is the same in every case — what changes is how much interpretation sits in front of it. You can switch models at any time, so if a phrasing doesn't land you can either say it more plainly or move up a tier.

Question 5

What's the philosophy behind it?

Accepted Answer

That the assistant should be an instrument that listens, not a collaborator with opinions. The friction in music production is rarely the idea — it's the distance between having the idea and executing it: the clicking, the menu diving, the four views you cross to do one thing. Tine exists to collapse that distance and then get out of the way. It augments the person with the taste rather than substituting for them, which is why it removes mechanical work and never creative work. If you're an artist who is sceptical of AI in music, we think that scepticism is largely correct, and we built this to be the thing it doesn't apply to.

Question 6

What principles is it built on?

Accepted Answer

Four, and they explain most of the design decisions. It performs the operation rather than explaining how to do it — an assistant that gives you instructions has just added a step. It asks instead of guessing: when a request is ambiguous, a clarifying question is always better than a confident wrong action on someone's work. It is deterministic wherever it can be, so the parts that must be reliable don't depend on a model being in a good mood. And it rewards knowing your craft: the more precisely you can name what you want in Ableton's own vocabulary, the more it does for you — there is no skill ceiling where it stops being useful.

Question 7

Will it wreck my project?

Accepted Answer

That is the objection that matters, and it drove the design. Tine resolves what you meant against a live read of your actual set instead of guessing, asks a clarifying question when a request is ambiguous rather than picking something, and asks for confirmation before heavy or destructive moves. Ordinary operations stay instant and undoable with Live's own undo.

Question 8

Does my project get uploaded anywhere?

Accepted Answer

No. Tine runs on your machine and talks to Live locally. You can run the language model entirely on your own GPU, in which case nothing leaves the computer at all. If you choose to use a hosted model instead, only the text of your request and a short description of the relevant part of your set are sent — never audio and never your project file.

Question 9

Do I have to talk to it?

Accepted Answer

No. Everything works typed. Voice matters most in the moments your hands are already committed — mid-take, holding a chord, a hand on a hardware knob — which is exactly where no keyboard shortcut can help you. But the microphone is optional and can be switched off entirely.

Question 10

Is it a subscription?

Accepted Answer

No. Tine is a one-time purchase with all 1.x updates included. You get a licence key that validates offline, so the app does not need to phone home to keep working.

Question 11

Is there a Mac version?

Accepted Answer

Not yet. Tine's native operations — headless save, render, and stem separation — are built directly against the Windows build of Live, and that work has to be redone for macOS rather than recompiled. Windows 10 and 11 are supported today.

Question 12

What happens when Ableton updates Live?

Accepted Answer

Tine's conversational and clip-editing features work through Live's official scripting interface and are stable across updates. The native operations are pinned to specific Live builds and are verified against each one, which is why the supported version is stated plainly on this page. If you update Live before Tine supports that build, the native features report that clearly instead of doing something unpredictable.

Question 13

How is this different from an MCP server that connects an AI to Ableton?

Accepted Answer

Those wire a chat model to Live's scripting interface, which is the easy part and works well on a clean, empty project. Tine's value is the two things that only show up on a real session: native operations Live's scripting interface cannot do at all, and a deterministic layer that resolves references, refuses ambiguous commands, and keeps an ordinary-sized model from doing damage in a 60-track project.

What you can ask for

Saving & exporting

Save without touching Live

Render audio in the background

Bounce a clip or track to audio

Separate a clip into stems

Turn audio into MIDI

Run an operation across a whole set

Start a new Live set

And the things it deliberately doesn’t do