Dictation
Hold the dictation keys, speak, let go. The text is typed into whatever window you're working in: an email, a document, a chat.
Try it in the demo →Dictate into any app. Record any call without a bot joining. Speech is turned into text on your computer, and there's no account to create.
Writing by voice and a record of your calls, in one search, without sending your speech away to be transcribed.
Hold the dictation keys, speak, let go. The text is typed into whatever window you're working in: an email, a document, a chat.
Try it in the demo →Click Start in a Teams, Zoom or browser call. Record keeps your microphone and the call's sound. No bot joins, and nobody gets an extra invite.
See a recorded meeting →The transcript is split by voice. Give a voice a name once and every line from it carries that name.
Name a voice →A short summary, actions with owners, what was discussed and what's still open. Each item opens the lines it came from, so you can check it, correct it and mark the notes reviewed. Made on your computer or by the AI provider you choose, with your own key.
Read the notes →Connect Outlook or Microsoft 365 to see today's meetings. A recording takes its meeting's title, and the people invited are offered as names for the voices.
One search across meeting titles, transcripts, notes and what you've dictated, opening the exact line.
Search the demo →What happens when you record a meeting. The listening and the transcript happen on your computer.
Your microphone and the call app's sound.
Speech turned into text, in time order.
Lines grouped by voice, named by you.
Summary, actions and topics, each linked to its lines.
A recorded meeting
Illustrative example. The meeting and the people are fictional. Open it in the demo.
The recording is deleted once its transcript is saved, unless you choose to keep meeting audio.
There's no Dscvry account. Apart from a check for updates, data leaves your computer only for things you turn on.
The full detail, including exactly what usage statistics contain: Record privacy policy.
The short answers. For anything else, ask us when you join the beta.
Windows, during the beta. Mac and Linux are planned. Tell us which computer you use when you sign up, and we'll invite you when Record is ready for it.
People who spend their day in calls and email and want a written record without sending their meetings to someone else's servers to be transcribed. It's free during the beta. We invite people in small groups and ask what works and what doesn't.
We've recorded Microsoft Teams, Zoom and calls in the browser. You can pick one app's sound or everything your computer plays. Record never starts on its own: you click Start. It can ask "Record this call?" when Teams starts playing sound, if you turn that on.
Yes. Tell the people in the call, and get any consent the law where you and they are requires. That's your responsibility under the beta terms.
English, Hindi and Telugu. English comes first; Hindi and Telugu are extra language packs you download in the app, each with its own dictation keys.
Dictation, transcripts, telling voices apart, notes made on your computer and search all work without a connection, once their models are downloaded. Notes from a cloud AI provider and the calendar need a connection.
Only if you turn it on. Both choices are in Settings, and both start off: anonymous usage statistics (counts and timings, such as how long dictation takes) and crash reports. They never include your words, audio, titles, names or files, and they go only to Dscvry. Exactly what is sent.
It works best when each person is on their own microphone, as in most calls. People sharing one microphone in a meeting room are often grouped as one voice today. That's one of the things we're working on.
Beta builds aren't yet signed with a code-signing certificate, so Windows SmartScreen asks before running the installer: choose "More info", then "Run anyway". Some managed work computers block unsigned apps completely; if yours does, tell us when you sign up. Updates inside the app are signed by Dscvry and checked before they install.
Yes. Speech recognition and AI notes make mistakes: a wrong word, the wrong speaker, an action missed or invented. Check anything you rely on against the transcript or the people involved. Every item opens the transcript lines it came from. You can correct an owner or wording, and if you make notes again, the earlier version is kept.
Click through the demo in your browser, then leave your details and we'll send you the installer when the next group of beta testers starts.