Browse documentation

Recording a Meeting

Start a capture with the Source Picker.

Start a recording from the Dashboard ("Start recording") or the Record item in the sidebar. This opens the Source Picker.

Using the Source Picker

  1. Name the meeting (optional) — a title is auto-generated otherwise.
  2. Pick a source:
    • • A screen (monitor) or a specific window — captures its audio and video.
    • Audio only — record microphone + system audio with no screen capture.
    • • A detected call — Teams, Zoom, Webex, Slack, or Discord — is flagged by name ("Zoom call detected") and recommended at the top.
  3. (Optional) Attach context & tools before you start:
    • Attach Context — pick Project Context documents to ground the assistant.
    • Enable MCP — turn on connected MCP servers/tools for this session.
  4. Click Record.
Your mic and system audio are captured as separate channels — which is how RecorderDesk reliably knows you versus everyone else, rather than guessing. No bot joins the call; capture happens on your machine and a red Recording Live indicator appears in the sidebar while the mic is live.

If something goes wrong mid-recording

  • The transcript is checkpointed as it goes. If you quit or the app is interrupted while recording, the words captured so far are kept — the session is journaled at stop, so an interrupted save can't lose the meeting.
  • Restart ASR Session — if live transcription stalls, this control in the transcript panel restarts the speech-recognition worker without ending your recording. RecorderDesk also supervises that worker and restarts it on its own before you need to.
  • • Prefer to open the mic deliberately? Turn on Start recordings muted under Settings → Audio & Microphone and unmute from the recording controls when you're ready.

Import instead of record

On the Dashboard you can Import media to bring an existing audio/video file into RecorderDesk and run the same transcription, diarization, and summary pipeline over it.

Picking a file first asks which language is spoken in it, defaulted to your current transcription language. That choice selects the on-device model used to transcribe the file — an English model would garble a Japanese recording — and it applies to that file only, leaving your default transcription language untouched. The recording keeps the language it was actually transcribed in, so the detail header reports it even if you change the setting later.

Was this page helpful? Get RecorderDesk