Transcription

RealtimeKit provides two transcription modes powered by Cloudflare Workers AI:

Mode	Model	Use Case
Real-time	Deepgram Nova-3	Live captions during meeting
End-of-meeting	Whisper Large v3 Turbo	Accurate offline transcription

Real-time transcription

Streams transcripts to participants as they speak using Deepgram Nova-3 via Cloudflare AI Gateway.

Enable via preset

Set transcription_enabled: true in the participant's preset:

{
  "name": "webinar_host",
  "transcription_enabled": true
}

Only participants with this flag will have their audio transcribed.

Configure

Pass ai_config.transcription when creating a meeting:

{
  "title": "Team Standup",
  "ai_config": {
    "transcription": {
      "language": "en-US",
      "keywords": ["RealtimeKit", "Cloudflare"],
      "profanity_filter": false
    }
  }
}

Option	Type	Default	Description
`language`	string	`en-US`	Language code for transcription
`keywords`	string[]	`[]`	Terms to boost recognition (names, jargon)
`profanity_filter`	boolean	`false`	Filter offensive language

Supported languages

Real-time transcription is powered by Deepgram Nova-3 on Workers AI.

Nova-3 on Workers AI supports the following languages for transcription:

Language	Code(s)
English	`en`, `en-US`, `en-AU`, `en-GB`, `en-IN`, `en-NZ`
Spanish	`es`, `es-419`
French	`fr`, `fr-CA`
German	`de`, `de-CH`
Hindi	`hi`
Russian	`ru`
Portuguese	`pt`, `pt-BR`, `pt-PT`
Japanese	`ja`
Italian	`it`
Dutch	`nl`

Use multi for automatic multilingual detection across all of the languages listed above.

If no language is specified, the model defaults to en-US. For best accuracy, explicitly set the language code matching your audio.

Consume transcripts

Client SDK

// Get all transcripts
const transcripts = meeting.ai.transcripts;

// Listen for new transcripts
meeting.ai.on("transcript", (data) => {
  if (data.isPartialTranscript) {
    // Interim result - speaker still talking
    updateLiveCaption(data.peerId, data.transcript);
  } else {
    // Final result
    appendToHistory(data);
  }
});

Transcript payload

{
  "id": "1a2b3c4d-5678-90ab-cdef-1234567890ab",
  "name": "Alice",
  "peerId": "4f5g6h7i-8j9k-0lmn-opqr-1234567890st",
  "userId": "uvwxyz-1234-5678-90ab-cdefghijklmn",
  "customParticipantId": "abc123xyz",
  "transcript": "Hello everyone",
  "isPartialTranscript": false,
  "date": "2024-08-07T10:15:30.000Z"
}

Field	Description
`isPartialTranscript`	`true` = interim (still speaking), `false` = final
`peerId`	Changes if participant rejoins
`userId`	Persistent participant ID
`customParticipantId`	Your custom ID from Add Participant API

End of meeting transcription

Generates transcripts after the meeting ends using Whisper Large v3 Turbo. Transcripts from all participants are consolidated into a unified timeline and delivered via webhook or REST API.

Enable end of meeting transcription

Set transcribe_on_end: true when creating a meeting:

{
  "title": "Team Standup",
  "transcribe_on_end": true
}

Use ai_config.transcription.language to set the transcript language. If transcribe_on_end is not set, end-of-meeting transcription is disabled and real time transcripts are generated.

Supported languages

Supports all languages in Whisper Large v3 Turbo. Use ISO 639-1 language codes, such as en, es, or fr.

Output formats

Format	Use Case
CSV	Spreadsheets, data analysis
SRT	Video subtitle files
VTT	Web video captions (`<track>` element)
JSON	Programmatic access

CSV example

Timestamp,Participant ID,User ID,Custom Participant ID,Participant Name,Transcript
2024-08-07T10:15:30.000Z,peer-123,user-456,cust-789,Alice,Hello everyone
2024-08-07T10:15:35.000Z,peer-234,user-567,cust-890,Bob,Hi Alice

JSON example

[
  {
    "startTime": 0,
    "endTime": 2.5,
    "sentence": "Hello everyone",
    "peerData": {
      "id": "peer-123",
      "userId": "user-456",
      "displayName": "Alice",
      "cpi": "cust-789"
    }
  }
]

Fetch transcripts

Webhook

Configure meeting.transcript event in webhooks:

{
  "event": "meeting.transcript",
  "meetingId": "meeting-123",
  "sessionId": "session-456",
  "transcriptDownloadUrl": "https://...",
  "transcriptDownloadUrlExpiry": "2024-08-14T10:15:30.000Z"
}

REST API

Refer to Fetch the complete transcript for a session.

curl -X GET "https://api.cloudflare.com/client/v4/accounts/{account_id}/realtime/kit/{app_id}/sessions/{session_id}/transcript" \
  -H "Authorization: Bearer {api_token}"

Transcripts are available for 7 days after meeting ends.