AI that runs on your own machine, and only if you install it.
Two families of tools that share one optional runtime. Pull a mix apart into stems, or generate, vary, and continue audio from a prompt, all inside the project and with your files staying on your disk. None of it ships in the base installer.
BS Roformer splits a mixed track into vocals, drums, bass, guitar, piano, and other. The results land as tracks in the project, ready to edit.
How separation worksACE-Step and Stable Audio 3 generate a clip from a prompt, create a variation of one you already have, continue it, or regenerate a selected range into a new result track.
The generation workflowsOne guided install from inside the app prepares a managed local runtime. No manual Python environment. Here is what it downloads and what it needs to run.
AI Tools setup guideThe base app stays lean. If you do not install the runtime, none of this is there.
Once the model files are on disk, generation and separation run without a connection.
Everything is decoded in full. If a model cannot run on your hardware, the app tells you rather than quietly producing something worse.
Separated parts arrive as tracks.
Hand BS Roformer a stereo mix and choose which stems you want. Vocals, drums, bass, guitar, piano, and other land in the arrangement as ordinary tracks that you can edit, mix, and render with the project. Use it for remixes, practice tracks, cleanup, or replacing a part.
Start a separationGenerate, vary, and continue from the timeline.
An AI track takes a prompt and optional lyrics and writes a fully decoded WAV into the session. Any existing audio clip can use variation, continuation or inpainting with a compatible model. Variation and inpainting create a new result track; continuation generates a tail. MiniMax Music 3 adds Lyrics + Style and Song Sections.
The models run through a diffusers pipeline. In our ACE-Step benchmark that path came in almost three times faster than the equivalent ComfyUI graph.
Read the benchmarkStyle prompt, optional lyrics, BPM, duration, key, seed. ACE-Step writes a WAV into the session.
Right-click a clip → AI Generation. A related version that keeps the source's identity.
Make a time selection over a clip and regenerate just that range to match what is around it.
Generate a tail that follows on from the selected clip, with prompt and length controls.
| Model | Family | What it does | Status |
|---|---|---|---|
| BS Roformer | Stem separation | Six-stem separation into new project tracks | Guided setup |
| ACE-Step | Generation | Text to Music, Lyrics + Style, variation, inpaint selection, continuation | Guided setup · Diffusers |
| Stable Audio 3 Medium | Generation | Text to Audio, variation, inpaint selection, continuation | Guided setup · Diffusers |
| MiniMax Music 3 | Generation | Lyrics + Style and Song Sections (structured songs) | Guided setup · Diffusers |
| Basic Pitch | Analysis | Audio to MIDI | Bundled model; inference in ONNX-enabled Windows and Linux builds |
The guided setup installs BS Roformer, ACE-Step, Stable Audio 3 Medium and MiniMax Music 3. Licence access and hardware support depend on the model. Intel Macs can run the base app, but the managed AI runtime currently targets Apple silicon.
Stable Audio 3 Medium
Request access on the Hugging Face model page and accept the Stability AI and Gemma licenses. Then choose Download and Set Up in the app, using a read token from your approved account or an existing Hugging Face login. OpenStudio downloads and converts the model automatically; local folder import remains optional. Allow extra disk space and time for conversion. This model supports text-to-audio, variation, selected-range replacement and continuation.
MiniMax Music 3
Accept the model license and choose Download and Set Up in the app. OpenStudio downloads the required Diffusers components from Hugging Face; a token is optional for this public model. Local folder import remains available. MiniMax supports lyrics and structured songs; source-audio editing uses another model. It needs substantial system RAM, and CPU offload trades GPU memory for more RAM and longer generation time.