Transcription Providers

Every MeetStream bot can transcribe its meeting. You pick the engine per request, under recording_config.transcript.provider:

{
  "meeting_link": "https://meet.google.com/abc-defg-hij",
  "bot_name": "Notetaker",
  "recording_config": {
    "transcript": {
      "provider": { "deepgram": { "model": "nova-3" } }
    }
  }
}

Post-call vs live

Post-call providers transcribe the recording after the meeting and produce a speaker-labeled transcript you fetch by transcript_id:

Streaming providers push transcript updates to your webhook during the meeting:

Streaming needs one extra field, live_transcription_required.webhook_url. See Live Transcription.

Platform support

jigsawstack_streaming is available on Google Meet. On Zoom and Teams, use deepgram_streaming, assemblyai_streaming, sarvam_streaming or meetstream_streaming. See the capability matrix for the rest of each platform.

How many providers you can use

• Post-call: list one provider in each request.
• Streaming: one provider runs. List only the one you want.
• Post-call plus streaming together: allowed. For example deepgram_streaming for live captions in your app and deepgram for a clean post-call transcript, in the same request.
• meetstream and jigsawstack cannot be combined in one request.
• meeting_captions can run alongside a streaming provider on Google Meet.

Who supplies the API key

Save your provider keys once in the dashboard under Integrations > Transcription. The Mia Transcribe card is the MeetStream engine and needs no key. The MeetStream engine costs $0.10 for each hour of recording, in addition to the bot rate. See pricing. Using a provider without a saved key fails at create time with 400, for example DeepgramAPIKey not configured in user profile.

Fetching post-call transcripts

1. Take the transcript_id from the Create Bot response (or from GET /api/v1/bots/{bot_id}/detail).
2. GET /api/v1/transcript/{transcript_id}/get_transcript returns the formatted transcript: an array of {speaker, transcript, start_time, end_time, words[]} with per-word confidence and speaker attribution. Add ?raw=true for the provider’s raw output. While processing it returns 202.
3. GET /api/v1/bots/{bot_id}/transcriptions lists every transcription run with status (Success / Processing / Failed) and download links (valid 1 hour).

Details: Post-call transcription.

Re-transcribe any recording

Recorded audio can be transcribed again with any post-call provider (different engine, different language) without re-recording:

curl -X POST "https://api.meetstream.ai/api/v1/bots/<BOT_ID>/transcribe" \
  -H "Authorization: Token <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{ "provider": { "assemblyai": {} } }'

• Exactly one provider per request: deepgram, assemblyai, jigsawstack, sarvam or meetstream.
• Returns 200 with a new transcript_id and processes in the background.
• 409 while the bot is still in the meeting. 404 if no audio was recorded (for example, the bot was never admitted). meetstream can be used once per bot (a second request returns 409). 402 if your balance is too low.

CLI: meetstream transcribe <bot-id> -p <provider>.