Be there.
Catch every word.

You need an Apple Silicon Mac (M1 or newer) running macOS 15 or later. Intel Macs are not supported by this download. You don’t need Xcode, Python, or an extra audio driver. Apple Translation requires macOS 26 or later and installed language packs; Apple’s meeting-notes provider also needs an eligible Mac with Apple Intelligence enabled.

Sottola captures system audio: the sound playing through your Mac from a call, video, or other app. It doesn’t join your meeting as a bot. Your own microphone is not captured by this system-audio path unless its sound is routed into system audio. Let participants know when you transcribe a conversation.

Local transcription, downloaded local translation, and local meeting notes run on your Mac after their models are prepared. The first download needs internet access. Speech and optional speaker models together are roughly 1 GB; local translation adds about 2.2 GB. The downloadable notes model needs additional storage. Optional cloud engines require an internet connection.

Meeting notes generate after you stop recording by default; you can turn automatic notes off or generate them manually. Choose Apple’s on-device model on an eligible Mac with macOS 26 or later, or download the local Qwen model. The same local providers can index folders you explicitly choose, so relevant file summaries can support your meeting notes. Context accepts UTF-8 text, Markdown, and source code, not PDF or Word files. Keep Sottola open until indexing and note generation finish. Neither feature automatically switches to a cloud model.

Language support depends on your selected speech and translation engines. Local Streaming does not support every language; Armenian, for example, is not supported. Select the spoken language in settings and check the result with your own audio. Optional speaker labels distinguish up to four anonymous voices; overlapping speech can be labeled “Multiple speakers.”

This release does not currently enforce a trial period or require a license key. A seven-day trial and paid licensing are planned for a future release. Pricing has not been announced.

With local engines selected, meeting audio and text are processed on your Mac. Model preparation connects to the model provider to download weights. Saved transcripts and context from folders you explicitly choose remain local. Selecting an optional OpenAI engine sends the relevant audio or text to that service. This website has no analytics scripts or advertising cookies; the hosting server keeps standard access logs.

Speech recognition, translation, and generated notes can make mistakes. Review important details against the transcript and source timestamps. Sottola’s notes help you follow up; they don’t send messages or create tasks on your behalf.