On-device speech
Download a speech model once, then transcribe meeting audio on the Mac without sending that content to a transcription service.
Sonari is designed so the meeting can be captured, transcribed, understood, searched, and summarized on your Mac. Privacy is visible in the workflow—not buried in a promise.
Each control answers a practical professional question: where did the content go, what produced this result, and what remains under my control?
Download a speech model once, then transcribe meeting audio on the Mac without sending that content to a transcription service.
Use a compatible local language model for summaries, actions, questions, and Ask Sonari answers.
Protect stored professional work and keep encrypted sharing packages distinct from ordinary exports.
Trace important outputs to the speaker, words, timestamp, and source audio instead of trusting a detached paragraph.
Inspect which functions stayed local and distinguish expected release, licensing, and model-download traffic from meeting-content processing.
Choose how long source audio remains and keep the transcript when policy allows it.
Sonari can use the network for licensing, signed software updates, model downloads, optional encrypted synchronization, calendar or email connections you enable, and privacy-safe stability reporting. These functions are separated from the local meeting-content path and should be disclosed clearly.
No. Sonari provides technical controls. Organizations remain responsible for consent, authorization, retention, supervision, security configuration, legal and regulatory obligations.
Sonari supports encrypted meeting packages and is designed so managed synchronization can store ciphertext rather than readable meeting content. Recipients and administrators must still be managed appropriately.
Yes. Ask Sonari is designed to cite the transcript with speaker, quotation, and timestamp so the user can check the result against the source.