Meetings may contain a language other than English, require automatic language detection, or include speakers switching between languages ("code-switching"). Provider support and configuration differ for each of these cases.

Code-switching is supported for both post-meeting and real-time transcription, but support varies by provider and model. Post-meeting models generally provide higher quality, so use them unless you need transcript data in real time.

Transcription providers that support code-switching

To use a third-party provider, create an account with that provider, obtain an API key, and add it to the Recall.ai transcription dashboard.

Real-time transcription code-switching configs

Add the appropriate provider configuration to your Create Bot request. Some providers are supported only for meeting bots. See the transcription support matrix for current availability.

Deepgram

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "deepgram",
    "deepgram": {
      "model": "nova-3",
      "language": "multi"
    }
  }
}

Gladia

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "gladia_v2",
    "gladia_v2": {
      "language_config": {
        "code_switching": true
      }
    }
  }
}

AssemblyAI

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "assembly_ai_v3",
    "assembly_ai_v3": {
      "speech_model": "universal-streaming-multilingual",
      "language_detection": true,
      "format_turns": true
    }
  }
}

Speechmatics

Speechmatics real-time transcription requires a supported bilingual or multi-language pack. For example, use cmn_en for Mandarin and English:

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "speechmatics",
    "speechmatics": {
      "language": "cmn_en"
    }
  }
}

AWS Transcribe

Set identify_multiple_languages to true for code-switching. Real-time transcription requires at least two candidate languages in language_options.

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "aws_transcribe",
    "aws_transcribe": {
      "identify_multiple_languages": true,
      "language_options": "en-US,es-US"
    }
  }
}

Choose only one of language_code, identify_language, or identify_multiple_languages. Include only one dialect per language in language_options.

Recall.ai Transcription

Recall.ai Transcription requires prioritize_accuracy for automatic language detection and code-switching. This mode delivers real-time transcript events with a 3-to-10-minute delay.

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "recallai",
    "recallai": {
      "mode": "prioritize_accuracy",
      "language_code": "auto"
    }
  }
}

ElevenLabs

Scribe v2 Realtime detects languages and code-switching automatically when language_code is omitted.

{
  "meeting_url": "...",
  "transcription_options": {
    "provider": "elevenlabs",
    "elevenlabs": {
      "model_id": "scribe_v2_realtime"
    }
  }
}

Rev does not currently support code-switching for real-time transcription.

Async transcription code-switching configs

Add the appropriate provider configuration to your Analyze Bot Media request.

Deepgram

{
  "deepgram_async_transcription": {
    "model": "nova-3",
    "language": "multi"
  }
}

Gladia

{
  "gladia_v2_async_transcription": {
    "language_config": {
      "code_switching": true
    }
  }
}

AssemblyAI

{
  "assemblyai_async_transcription": {
    "speech_models": ["universal-2"],
    "language_detection": true,
    "language_detection_options": {
      "code_switching": true
    }
  }
}

Speechmatics

For automatic code-switching across Speechmatics-supported languages, use the Melia 1 model:

{
  "speechmatics_async_transcription": {
    "transcription_config": {
      "model": "melia-1",
      "language": "multi"
    }
  }
}

You can instead select a supported bilingual or multi-language pack for the Standard or Enhanced model:

{
  "speechmatics_async_transcription": {
    "language": "cmn_en"
  }
}

AWS Transcribe

{
  "aws_transcribe_async_transcription": {
    "identify_multiple_languages": true,
    "language_options": ["en-US", "es-US"]
  }
}

language_options is optional for AWS Transcribe async multi-language identification. Choose only one of language_code, identify_language, or identify_multiple_languages.

Recall.ai Transcription

{
  "recallai_async_transcription": {
    "language_code": "auto"
  }
}

ElevenLabs

Scribe v2 detects languages and code-switching automatically when language_code is omitted.

{
  "elevenlabs_async_transcription": {
    "model_id": "scribe_v2"
  }
}

Rev

Rev supports English/Spanish code-switching for async transcription:

{
  "rev_async_transcription": {
    "language": "en/es"
  }
}

FAQ

What is the difference between language detection and code-switching?

Language detection selects a single language for a recording or stream. Code-switching allows the transcript to change languages when the spoken language changes.

What languages are supported by each provider?

Provider support changes over time and may vary by model and transcription workflow. Use the provider documentation linked in the table above to confirm current language and model support.