Meeting BaaSMeeting BaaS
MeetingBaas APIMeetingBaas API v2Typescript SDKMCP ServersSpeaking BotsTranscript SeekerSelf Hosting

Transcription

Complete guide to transcription features, custom parameters, and BYOK

Meeting BaaS v2 provides powerful transcription capabilities with support for custom parameters, multiple providers, and Bring Your Own Key (BYOK) options.

Overview

Transcription in v2 offers:

  • Multiple Providers: Currently supports Gladia, with Assembly AI and Deepgram coming soon
  • BYOK Support: Use your own transcription provider API keys to save on token costs
  • Custom Parameters: Configure LLM summaries, translation, language detection, and more
  • Raw & Processed Output: Access both raw provider responses and standardized transcriptions
  • Transcription IDs: Track transcription jobs for BYOK users

Enabling Transcription

To enable transcription for a bot, include transcription_config in your bot creation request:

{
  "meeting_url": "https://meet.google.com/...",
  "bot_name": "AI Notetaker",
  "transcription_enabled": true,
  "transcription_config": {
    "provider": "gladia",
    "api_key": null,
    "custom_params": null
  }
}

Basic Configuration

Required Fields:

  • transcription_enabled: Set to true to enable transcription
  • transcription_config.provider: Currently "gladia" (default). Assembly AI and Deepgram support coming soon.

Optional Fields:

  • transcription_config.api_key: Your transcription provider API key (for BYOK - see below)
  • transcription_config.custom_params: Custom parameters for advanced features (see below)

Transcription Providers

Current Provider

Gladia (Default)

  • High-accuracy transcription
  • Speaker diarization
  • Multi-language support
  • Advanced features (summarization, translation, etc.)

Coming Soon

  • Assembly AI: Additional transcription provider option
  • Deepgram: Additional transcription provider option

Provider selection will be available via the provider field in transcription_config.

Bring Your Own Key (BYOK)

Using your own transcription provider API key can significantly reduce token costs. When you provide your own key:

  • Token Savings: Transcription tokens are reduced from 0.25 tokens/hour to 0.05 tokens/hour
  • Provider Billing: You're billed directly by the transcription provider
  • Full Control: Manage your own provider account and usage

Requirements

BYOK transcription is available on Pro plans and above. Pay-as-you-go plans use the platform's transcription keys.

Setting Up BYOK

  1. Get Your API Key: Obtain an API key from your transcription provider (e.g., Gladia)
  2. Include in Request: Add the API key to transcription_config.api_key:
{
  "transcription_enabled": true,
  "transcription_config": {
    "provider": "gladia",
    "api_key": "your-gladia-api-key-here",
    "custom_params": null
  }
}
  1. Track Jobs: Use transcription_ids in bot details and webhooks to track your provider jobs

Custom Parameters

v2 supports advanced transcription features through custom parameters. These are provider-specific options that enhance transcription capabilities.

For complete documentation on all available custom parameters, see the Gladia API Reference.

Available Custom Parameters

Summarization

Generate AI-powered meeting summaries:

{
  "transcription_config": {
    "provider": "gladia",
    "custom_params": {
      "summarization": true,
      "summarization_config": {
        "type": "general"  // or "bullet_points", "concise"
      }
    }
  }
}

Summary Types:

  • general: General meeting summary
  • bullet_points: Bullet-point format
  • concise: Concise summary

Translation

Translate transcriptions to multiple languages:

{
  "transcription_config": {
    "provider": "gladia",
    "custom_params": {
      "translation": true,
      "translation_config": {
        "target_languages": ["es", "fr", "de"],
        "model": "enhanced",  // or "base"
        "match_original_utterances": true,
        "lipsync": true,
        "context_adaptation": true
      }
    }
  }
}

Translation Options:

  • target_languages: Array of ISO 639-1 language codes (e.g., ["es", "fr"])
  • model: "base" (default) or "enhanced" for better quality
  • match_original_utterances: Match translated utterances to original timing
  • lipsync: Enable lip-sync for video
  • context_adaptation: Adapt translation to context

Language Detection

Force specific languages or enable automatic detection:

{
  "transcription_config": {
    "provider": "gladia",
    "custom_params": {
      "language_config": {
        "languages": ["en", "es"],  // Force specific languages
        "detect_language": true  // Enable automatic detection
      }
    }
  }
}

Subtitles

Generate subtitles in multiple formats:

{
  "transcription_config": {
    "provider": "gladia",
    "custom_params": {
      "subtitles": true,
      "subtitles_config": {
        "formats": ["srt", "vtt"],
        "minimum_duration": 0.5,
        "maximum_duration": 5,
        "maximum_characters_per_row": 42,
        "maximum_rows_per_caption": 2,
        "style": "default"  // or "compliance"
      }
    }
  }
}

Custom Vocabulary

Improve accuracy for domain-specific terms:

{
  "transcription_config": {
    "provider": "gladia",
    "custom_params": {
      "custom_vocabulary": [
        "MeetingBaaS",
        "API",
        "webhook"
      ],
      "custom_vocabulary_config": {
        "vocabulary": [
          {
            "value": "MeetingBaaS",
            "intensity": 0.8,
            "pronunciations": ["meeting-baas", "meeting-bass"]
          }
        ],
        "default_intensity": 0.7
      }
    }
  }
}

Additional Features

Other available custom parameters:

  • Moderation: Content moderation and filtering
  • Named Entity Recognition: Extract names, organizations, locations
  • Sentiment Analysis: Analyze sentiment of utterances
  • Chapterization: Automatically create meeting chapters
  • Name Consistency: Maintain consistent speaker names
  • Custom Spelling: Custom spelling dictionary
  • Structured Data Extraction: Extract structured data using class definitions
  • Audio to LLM: Apply LLM prompts to transcription output
  • Punctuation Enhanced: Enhanced punctuation accuracy

For complete documentation on all custom parameters and their configuration options, see the Gladia API Reference. You can also check the Meeting BaaS API Reference for our API schema.

Transcription Output

v2 provides two types of transcription files:

Raw Transcription (raw_transcription.json)

Contains the complete, unmodified response from the transcription provider:

  • Includes: All custom parameters (LLM summaries, translations, metadata)
  • Format: Provider-specific structure
  • Use Case: Access advanced features like summaries, translations, custom metadata
  • Note: Presented as an array without time duration offsets or speaker diarization. Best used alongside output_transcription.json.

Example Structure:

{
  "bot_id": "uuid",
  "transcriptions": [
    {
      "transcription": {
        "utterances": [...],
        "summary": "Meeting discussed Q4 goals...",
        "languages": ["en", "es"],
        "metadata": {...}
      }
    }
  ]
}

Output Transcription (output_transcription.json)

Standardized format across all providers:

  • Format: Consistent structure regardless of provider
  • Features:
    • Speaker diarization with participant names
    • Timestamps adjusted for multi-chunk recordings
    • Standardized utterance format
  • Use Case: General transcription processing and display

Example Structure:

{
  "result": {
    "utterances": [
      {
        "start": 0.5,
        "end": 2.1,
        "text": "Hello everyone",
        "speaker": "John Doe",
        "language": "en"
      }
    ]
  }
}

Accessing Transcriptions

Transcriptions are available via presigned S3 URLs in:

  1. Bot Details (GET /v2/bots/:bot_id): Artifacts array includes transcription URLs
  2. Webhooks (bot.completed): Transcription URLs in webhook payload
  3. Callbacks: Same URLs as webhooks

Presigned URLs

  • Validity: 4 hours from generation
  • Security: Time-limited access to transcription files
  • Download: Fetch and store transcriptions promptly

Example Webhook Payload:

{
  "event": "bot.completed",
  "data": {
    "bot_id": "uuid",
    "raw_transcription": "https://s3.amazonaws.com/.../raw_transcription.json",
    "transcription": "https://s3.amazonaws.com/.../output_transcription.json",
    "transcription_ids": ["gladia-job-12345"],
    "transcription_provider": "gladia"
  }
}

Transcription IDs

For BYOK users, transcription_ids provides an array of provider job IDs:

  • Purpose: Track your own transcription jobs with the provider
  • Error Correlation: Match transcription errors to specific provider job IDs
  • Multi-Chunk Support: Each audio chunk gets its own provider ID
  • Available In: Bot details and webhook payloads

Example:

{
  "transcription_ids": ["gladia-job-12345", "gladia-job-12346"],
  "transcription_provider": "gladia"
}

Token Consumption

Transcription token consumption depends on whether you use BYOK:

With Platform Key (Default)

  • Recording: 1 token/hour
  • Transcription: 0.25 tokens/hour
  • Total: ~1.25 tokens/hour

With BYOK

  • Recording: 1 token/hour
  • BYOK Transcription: 0.05 tokens/hour
  • Total: ~1.05 tokens/hour

Note: Custom parameters (summarization, translation, etc.) may incur additional costs from the transcription provider when using BYOK. Check your provider's pricing.

Error Handling

If transcription fails:

  • Error Code: TRANSCRIPTION_FAILED in bot status
  • Token Charging: Recording and streaming tokens are charged, but transcription tokens are not
  • Retry: Use the re-transcribe endpoint to retry transcription
  • Webhook: bot.failed webhook includes error details

See Error Codes for complete error information.

Best Practices

  1. Use BYOK for Cost Savings: If you have high transcription volume, BYOK can significantly reduce costs
  2. Download Promptly: Presigned URLs expire after 4 hours - download transcriptions quickly
  3. Store Long-Term: If you need transcriptions long-term, download and store them in your own storage
  4. Use Raw Transcription: Access LLM summaries, translations, and custom features from raw transcription
  5. Track Transcription IDs: For BYOK users, use transcription_ids to correlate with provider jobs
  6. Handle Errors: Implement retry logic for transcription failures

Examples

Basic Transcription

curl -X POST "https://api.meetingbaas.com/v2/bots" \
  -H "x-meeting-baas-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_url": "https://meet.google.com/abc-defg-hij",
    "bot_name": "Notetaker",
    "transcription_enabled": true,
    "transcription_config": {
      "provider": "gladia"
    }
  }'

BYOK with Summarization

curl -X POST "https://api.meetingbaas.com/v2/bots" \
  -H "x-meeting-baas-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_url": "https://meet.google.com/abc-defg-hij",
    "bot_name": "AI Notetaker",
    "transcription_enabled": true,
    "transcription_config": {
      "provider": "gladia",
      "api_key": "your-gladia-api-key",
      "custom_params": {
        "summarization": true,
        "summarization_config": {
          "type": "bullet_points"
        }
      }
    }
  }'

Multi-Language Translation

curl -X POST "https://api.meetingbaas.com/v2/bots" \
  -H "x-meeting-baas-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "meeting_url": "https://meet.google.com/abc-defg-hij",
    "bot_name": "Multilingual Notetaker",
    "transcription_enabled": true,
    "transcription_config": {
      "provider": "gladia",
      "custom_params": {
        "translation": true,
        "translation_config": {
          "target_languages": ["es", "fr", "de"],
          "model": "enhanced"
        },
        "language_config": {
          "detect_language": true
        }
      }
    }
  }'

Next Steps

Edit on GitHub

On this page

Overview
Enabling Transcription
Basic Configuration
Transcription Providers
Current Provider
Coming Soon
Bring Your Own Key (BYOK)
Requirements
Setting Up BYOK
Custom Parameters
Available Custom Parameters
Summarization
Translation
Language Detection
Subtitles
Custom Vocabulary
Additional Features
Transcription Output
Raw Transcription (raw_transcription.json)
Output Transcription (output_transcription.json)
Accessing Transcriptions
Presigned URLs
Transcription IDs
Token Consumption
With Platform Key (Default)
With BYOK
Error Handling
Best Practices
Examples
Basic Transcription
BYOK with Summarization
Multi-Language Translation
Next Steps