by pyannote
speaker-diarization-3.1 from pyannote is an open-weight model — the weights are published and can be downloaded, self-hosted, and fine-tuned. It falls under the Speech recognition category, and its license permits commercial use.
A short, permissive license: use, modify, and distribute freely — including in commercial and closed-source products — as long as you keep the copyright and license notice.
Plain-language guidance, not legal advice — always confirm against the model’s own license before you rely on it.