Changelog
Voice dictation starts listening the moment you press
We've made voice dictation start listening the instant you press the mic, in the dashboard and in the iPhone app. Before, your first words could go missing while the connection set up. Now everything you say from the press on is captured, and the transcript is more accurate: we moved it to AssemblyAI's newest live transcription model, which recognizes 32 languages and switches between them on its own. Read why talking to agents beats typing.
How it works
- Recording starts on the press. Your words are held on your device while the transcription connection opens, then sent in order, so a quick "summarize this thread" is never cut short.
- Product and integration names come out right. Dictation is told to expect geni, General Input, and the names of the tools you connect, like HubSpot, QuickBooks, Airtable, and Google Sheets, before the cleanup pass ever sees them.
- One permission prompt, on your first press. Hovering the mic no longer makes the browser ask for microphone access. If access is blocked, the mic shows a crossed-out icon and its menu says how to allow it.
- 32 languages. Up from six, detected automatically, including mid-sentence switches.
Also shipped
- Dictation is now billed for the time the mic is listening, at about $0.38 per hour, prorated per second.
- The iPhone changes arrive with the next app update.