AI models

Two features in Nota use AI models. They’re optional downloads, and they run on your computer — the audio you give them never leaves it.

Model	Used by	Size	By
htdemucs	Separate Stems — splits a mix into drums, bass, vocals and the rest	about 175 MB	Meta AI (Demucs v4), MIT
basic-pitch	Convert to MIDI — hears every note, chords included	under 1 MB	Spotify, Apache-2.0

Both run on ONNX Runtime, which comes with the first model you install and goes with the last one you remove.

Installing and removing

Settings ▸ Downloads ▸ AI Models: Install downloads a model from where its authors publish it and checks it against the checksum pinned in Nota; Remove deletes it (the feature then offers to download it again); Source opens the model’s project page.

You don’t have to come here first — Separate Stems… and the Convert menu offer the download the first time you need it. Downloads keep going after you close Settings; their progress moves to the main window’s status line.

Note

If ONNX Runtime has no build for your computer, the page says so and the models can’t run there.