Record a meeting with live transcription
Live transcription shows speech during a meeting and creates meeting minutes with a summary, decisions and ToDos after recording stops. For transcripts already received from Zoom, see Meetings and Zoom imports.
Before you start
- Use desktop Chrome or Edge. Unsupported devices show why recording is unavailable.
- An administrator must enable meeting recording in workspace settings. You need permission to operate in the target Room.
- Confirm every participant’s consent to recording and transcription. Starting recording confirms that you have obtained their consent. Screen images are not saved.
- The transcription model downloads on first use. Prepare it before the meeting.
Choose Japanese or English before recording. English is available when enabled for your environment. If an English session cannot start, check the displayed availability message.
Start recording
- Open Terasu AI in the target Room
- Open the input’s add menu and choose the meeting-recording action
- Check the available language choices and consent notice, then start recording
- Choose the meeting type in the separate recorder window
- Review the browser’s microphone and screen-sharing permissions and start capture
Choose the audio source
| Meeting location | Input | What to check |
|---|---|---|
| Browser meeting | Browser tab | Select the meeting tab and enable tab audio sharing |
| Desktop meeting app | Desktop app | Select the screen and enable system audio sharing |
| In-person meeting | Microphone only | Allow microphone access. This input does not separate your voice from other speakers |
On macOS, desktop-app capture may require allowing your browser under System Settings → Privacy & Security → Screen Recording, followed by a browser restart. Without permission, the sharing picker may show only tabs.
Headphones reduce overlap between your microphone and the other participants’ audio. Check the recorder’s input-level meters.
Stop and open the minutes
Transcript text appears as speech is processed. You may minimize the recorder window; closing it stops recording.
Choose Stop and keep the window open while the remaining transcript is sent. The Room then creates the minutes. Use the open-minutes action to review the summary, decisions, ToDos and transcript. No minutes are created when there is no finalized transcript. Recording also stops at the four-hour limit.
Troubleshooting
| Symptom | Action |
|---|---|
| Recorder window does not open | Allow browser pop-ups |
| Your voice is missing | Check microphone permission and your input-level meter |
| Other voices are missing | Share again with tab audio or system audio enabled |
| No text after one minute | Check input levels and choose the correct audio source |
| Model preparation fails | Check connectivity and available device storage, then use Retry |
| Recording ends with a failure | Read the Room card. A saved partial transcript is different from a session with no transcript and no minutes |
Storage and retention depend on workspace settings. Live capture and post-meeting Zoom import are separate workflows.
Start a recording session through MCP
MCP additions not included in the first release:
create_meeting_recording、prepare_create_meeting_recording. These additions are planned for a later release. Existing screen operations remain available.
Use prepare_create_meeting_recording then create_meeting_recording to create an unstarted session in the target Room. This requires files:write, current internal Workspace membership, and native recording permission. Participant consent, recording policy, and the monthly usage limit apply.
The same user opens the returned appPath in their authenticated browser, chooses the audio source, and reviews browser permissions in the existing recorder window. Successful MCP creation has not started audio capture. Use appPath first: pressing the Room’s start button again creates a different session.
The handover is available for the creator’s unclaimed session within four hours of creation. Reopening the URL does not extend that deadline. For an active, cancelled, or expired session, read the message and use the Room return link to inspect its state. English MCP creation requires general English availability; limited-cohort users create through the existing browser screen. Current availability also applies when opening a legitimately created English session.
Review the preparation and submit identical input, operation key, and confirmation within ten minutes. Inspect get_mcp_operation without automatically replaying unknown outcomes. See MCP operations and limits for inputs and budgets.