Skip to main content

​
What can you do with it?

The /transcription-deepgram command enables you to transcribe audio and video files using Deepgram’s advanced AI transcription service. Perfect for converting meetings, interviews, podcasts, and video content into text with features like speaker identification, sentiment analysis, topic extraction, and automatic formatting.

​
How to use it?

​
Basic Command Structure

/transcription-deepgram [audio-file or video-file] [options]

​
Parameters

Required:
  • fileUrls - The audio or video file to transcribe (URL or uploaded file). Supports both audio and video formats.
Optional:
  • model - Deepgram model to use: nova-3 (default, recommended), nova-2, enhanced, or base
  • languageCode - Language code (e.g., “en”, “es”, “fr”). Default: auto-detect
  • enableAutomaticPunctuation - Add punctuation automatically (default: true)
  • enableDiarization - Identify different speakers (default: false)
  • diarizationSpeakerCount - Number of expected speakers (required if diarization is enabled)
  • enableUtterances - Segment transcription by speaker turns (default: false)
  • enableParagraphs - Format output into paragraphs (default: false)
  • enableSummary - Generate AI summary of content (default: false)
  • enableTopics - Extract key topics discussed (default: false)
  • enableIntents - Identify user intents (default: false)
  • enableSentiment - Analyze sentiment (positive/negative/neutral) (default: false)
  • customTopics - Custom topics to detect (array of strings)
  • customIntents - Custom intents to detect (array of strings)
  • collectionId - File store resource ID for storing results (use /filestore slash command to select, default: multimedia artifact collection)
  • triggerUrls - Webhook URLs to notify when processing completes
  • outputFileNames - Output file names for async mode (can be .json or .txt format)

​
Response Format

Async Response (for large files or when background processing is needed): 
{
  "success": true,
  "capability": "transcription",
  "provider": "deepgram",
  "async": true,
  "collectionId": "collection-id",
  "responseId": "transcription-processing-queue-[TIMESTAMP]-[RANDOM]",
  "status": "queued",
  "createdAt": "ISO timestamp",
  "results": [
    {
      "inputFileName": "original-file-name",
      "outputFileName": "transcribe-with-speech-model-[RANDOM].json",
      "fileId": "file-id",
      "signedUrl": "https://skills-stage.pinkfish.ai/files/[COLLECTION_ID]/[OUTPUT_FILE]?signed-key=..."
    }
  ],
  "message": "Deepgram transcription processing started in background. Results will be saved to file storage when complete."
}
Sync Response (for immediate results): 
{
  "success": true,
  "async": false,
  "collectionId": "collection-id",
  "capability": "transcription",
  "provider": "deepgram",
  "model": "nova-3",
  "languageCode": "auto-detect",
  "totalFiles": 1,
  "successfulFiles": 1,
  "failedFiles": 0,
  "totalAudioDuration": 2.5,
  "results": [
    {
      "fileName": "sample.wav",
      "inputFileName": "sample.wav",
      "inputMimeType": "application/octet-stream",
      "transcription": "Full transcription text here...",
      "wordCount": 100,
      "audioDuration": 2.5,
      "fileId": "file-id"
    }
  ]
}

​
Examples

​
Basic Transcription

/transcription-deepgram
fileUrls: meeting-recording.mp3
Transcribes an audio file with default settings (auto punctuation, nova-3 model).

​
Meeting Transcription with Speaker Identification

/transcription-deepgram
fileUrls: team-meeting.mp3
enableDiarization: true
diarizationSpeakerCount: 3
enableUtterances: true
enableParagraphs: true
Creates a transcript with speaker identification, formatted by speaker turns and paragraphs.

​
Video Transcription with Analysis

/transcription-deepgram
fileUrls: presentation-video.mp4
enableSummary: true
enableTopics: true
enableSentiment: true
enableParagraphs: true
Transcribes a video file and provides a summary, key topics, sentiment analysis, and formatted paragraphs.

​
Customer Service Analysis

/transcription-deepgram
fileUrls: customer-call.mp3
enableDiarization: true
enableIntents: true
enableSentiment: true
customIntents: ["complaint", "question", "request"]
Transcribes a customer service call with speaker identification, intent detection, sentiment analysis, and custom intent categories.

​
Async Processing for Large Files

/transcription-deepgram
fileUrls: long-podcast.mp3
outputFileNames: ["transcript.json", "transcript.txt"]
Starts background processing for a large file and saves results to specified output files when complete.

​
Quick Sync Transcription

/transcription-deepgram
fileUrls: short-clip.wav
Quickly transcribes a short audio clip and returns immediate results.

​
When to Use Async vs Sync

  • Use ASYNC for:
    • Large files (more than 5 minutes)
    • Multiple files
    • When you need to continue working while transcription processes
    • When you want to specify custom output file names
  • Use SYNC for:
    • Quick transcriptions
    • Short audio clips (less than 5 minutes)
    • When immediate results are needed
    • Simple, single-file transcriptions

​
Usage Recommendations

​
For Meeting Transcriptions

Enable enableDiarization=true, enableUtterances=true, and enableParagraphs=true to get speaker-labeled, well-formatted transcripts.

​
For Content Analysis

Enable enableSummary=true, enableTopics=true, and enableSentiment=true to extract insights from audio/video content.

​
For Customer Service

Enable enableIntents=true, enableSentiment=true, and enableDiarization=true to analyze customer interactions.

​
For Quick Transcripts

Use default settings with enableAutomaticPunctuation=true for fast, accurate transcription without additional features.

​
Supported Models

  • nova-3 (default, recommended) - Latest and most accurate model
  • nova-2 - Previous generation, high accuracy
  • enhanced - High accuracy for challenging audio conditions
  • base - Faster, cost-effective option for clear audio

​
Common Use Cases

  • “Transcribe this meeting with speaker identification” → Use enableDiarization=true and enableUtterances=true
  • “Get a summary of this audio” → Use enableSummary=true
  • “What topics are discussed?” → Use enableTopics=true
  • “Analyze the sentiment” → Use enableSentiment=true
  • “Format it nicely with paragraphs” → Use enableParagraphs=true

​
Notes

Supported Formats:
  • Audio: MP3, WAV, M4A, FLAC, OGG, AAC, AIFF
  • Video: MP4, MOV, AVI, and other common video formats
Language Support:
  • Automatic language detection
  • Over 50 languages supported
  • Specify language code for better accuracy
Processing:
  • Async mode creates a placeholder file that updates when processing completes
  • Use the signedUrl from the response to monitor processing status
  • Results saved to the specified collection (default: multimedia artifact collection)