For clean Markdown of any page, append .md to the page URL. For a complete documentation index, see https://docs.meetstream.ai/guides/transcription-recordings/llms.txt. For full documentation content, see https://docs.meetstream.ai/guides/transcription-recordings/llms-full.txt. # MeetStream Guide: Per-Participant Audio Streams This guide explains how to capture **individual audio recordings per participant** in a meeting. Instead of a single mixed audio track, you receive a separate WebM audio file for each speaker. Supported platforms: **Google Meet, Zoom**. --- ## 1) What you need 1. A MeetStream API key (refer to [Dashboard Setup](/guides/get-started/dashboard-setup)). 2. A meeting link (Google Meet or Zoom). 3. Set `audio_separate_streams: true` in your Create Bot request. --- ## 2) Create a bot with per-participant audio Use the **Create Bot** endpoint: https://docs.meetstream.ai/api-reference/ap-is/bot-endpoints/create-bot ### Minimal example ```bash curl -X POST "https://api.meetstream.ai/api/v1/bots/create_bot" \ -H "Authorization: Token " \ -H "Content-Type: application/json" \ -d '{ "meeting_link": "", "bot_name": "RecorderBot", "audio_separate_streams": true }' ``` ### Full example with recording config ```json { "meeting_link": "https://meet.google.com/abc-defg-hij", "bot_name": "RecorderBot", "audio_separate_streams": true, "recording_config": { "transcript": { "provider": { "deepgram": { "model": "nova-3", "language": "en" } } } }, "automatic_leave": { "everyone_left_timeout": 5 } } ``` ### Parameter reference | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | `audio_separate_streams` | boolean | `false` | Enable per-participant audio capture | > **Tip:** You can also pass `audio_separate_streams` nested inside `recording_config`: > ```json > "recording_config": { > "audio_separate_streams": true > } > ``` > The top-level parameter takes precedence if both are set. --- ## 3) What happens during the meeting Once the bot joins, it automatically: - Captures each participant's audio as a **separate file**. - Records up to **16 concurrent speaker streams**. - Continues recording throughout the meeting -- no configuration needed per participant. The bot also continues to produce the standard **mixed audio** file (`audio.wav`) regardless of whether `audio_separate_streams` is enabled. > **Note on platform behaviour:** The mechanism for capturing per-participant audio differs between Zoom and Google Meet and affects the isolation level of each file. See [Section 7](#7-per-participant-audio-by-platform) for details. --- ## 4) Retrieve per-participant audio streams After the bot leaves and audio processing completes, call: ```bash curl -X GET "https://api.meetstream.ai/api/v1/bots//get_audio_streams" \ -H "Authorization: Token " ``` ### Response when processing is complete ```json { "bot_id": "f923cd21-da86-4d77-b37b-0707d37751c8", "audio_status": "Success", "audio_streams_available": true, "participants": [ { "participant_name": "John Doe", "streams": [ { "stream_id": "JohnDoe_abc123", "segments": [ { "segment_index": 0, "url": "https://s3.amazonaws.com/...?X-Amz-Signature=...", "filename": "John_Doe_abc123_0.webm", "duration_seconds": 125.5, "sample_rate": 48000, "channels": 1, "codec": "opus" } ] } ] }, { "participant_name": "Jane Smith", "streams": [ { "stream_id": "JaneSmith_def456", "segments": [ { "segment_index": 0, "url": "https://s3.amazonaws.com/...?X-Amz-Signature=...", "filename": "Jane_Smith_def456_0.webm", "duration_seconds": 118.3, "sample_rate": 48000, "channels": 1, "codec": "opus" } ] } ] } ], "summary": { "total_participants": 2, "total_segments": 2 } } ``` ### Response when processing is still in progress ```json { "audio_status": "in_progress", "message": "Bot is still in the meeting. Audio streams will be available after the bot leaves." } ``` HTTP status **202** is returned when the bot has not yet left the meeting. Poll again after the bot exits. ### Response when no audio streams are available ```json { "bot_id": "f923cd21-da86-4d77-b37b-0707d37751c8", "audio_status": "Success", "audio_streams_available": false, "message": "No per-participant audio streams available for this bot.", "participants": [] } ``` This is returned when `audio_separate_streams` was not enabled, or the meeting ended before any speech was detected. --- ## 5) Download the audio files Each segment in the response contains a `url` field -- a **presigned S3 URL** that allows direct download without additional authentication. ```bash # Download a participant's audio curl -o "John_Doe.webm" "" ``` > **Important:** Presigned URLs expire after **10 minutes**. If a URL has expired, call `get_audio_streams` again to get fresh URLs. ### File format | Property | Value | |----------|-------| | Container | WebM | | Audio codec | Opus | | Bitrate | 48 kbps | | Sample rate | 48,000 Hz | | Channels | Mono | --- ## 6) Understanding segments Each participant has one segment (`segment_index: 0`) covering the full duration of their speech in the meeting. Silence gaps between speech periods are preserved with actual silence, so the file's timeline aligns with the real meeting timeline. Use `duration_seconds` to determine how long a participant was speaking. --- ## 7) Per-participant audio by platform The quality of speaker isolation differs by platform due to how each platform delivers audio. | Platform | Isolation | Mechanism | |----------|-----------|-----------| | **Zoom** | Full isolation | Zoom SDK provides a dedicated raw PCM stream per participant. Each file contains only that participant's microphone audio. | | **Google Meet** | Partial isolation | Up to 3 concurrent speaker streams are captured via WebRTC CSRC demuxing. Audio is attributed to whoever is speaking at each moment. | > **Zoom** produces the cleanest per-participant files. **Google Meet** files contain the meeting's mixed audio during the periods when that participant was the active speaker. --- ## 8) Using both mixed and per-participant audio Per-participant audio and the standard mixed audio are always captured simultaneously. You do not need to choose one or the other. ```json { "meeting_link": "", "audio_separate_streams": true } ``` This produces: - A **mixed audio WAV** (all participants combined) -- retrieve via the **Get Bot Audio** endpoint (`/api/v1/bots//get_audio`). - **Per-participant WebM files** -- retrieve via the **Get Audio Streams** endpoint (`/api/v1/bots//get_audio_streams`). --- ## 9) Using per-participant audio with per-participant video Both flags can be enabled together: ```json { "meeting_link": "", "audio_separate_streams": true, "video_separate_streams": true } ``` This produces separate audio and video files per participant. The files are not muxed together -- audio and video are delivered as independent files. Match them by `participant_name` across the two API responses. --- ## 10) Webhook notifications If you have webhooks configured, you will receive an `audio.processed` event when audio processing (including per-participant stream generation) completes. Poll `get_audio_streams` until `audio_streams_available` is `true` if you prefer polling over webhooks. --- ## 11) Troubleshooting - **`audio_streams_available: false` with `audio_status: "Success"`?** - Confirm that `audio_separate_streams: true` was set when the bot was created. - The meeting may have ended before any speech was detected. - **Missing a participant's audio?** - The bot captures up to **16 concurrent speaker streams**. Meetings with more than 16 active speakers may result in some participants not being captured. - Participants who never spoke will not appear in the response. - **Getting status `in_progress`?** - The bot is still in the meeting. Audio streams are generated after the bot exits. Wait for the bot to leave and poll again. - **Presigned URL returns 403 Forbidden?** - The URL has expired (10-minute lifetime). Call `get_audio_streams` again for fresh URLs. - **Audio file sounds like the whole meeting, not just one person?** - This is expected on **Google Meet** -- see [Section 7](#7-per-participant-audio-by-platform). The file contains the meeting audio during that participant's active speaking windows. On **Zoom**, files are fully isolated.