Transcribe Voice Notes

Transcribe WhatsApp voice notes with automatic correction of commonly mis-transcribed mining terminology.

When to Use This Skill

• You have WhatsApp voice notes to process
• You want automatic transcription with mining-site name corrections
• You need structured task creation from voice note content

Usage

Via slash command:

/transcribe-voice-notes

Via Python (if running independently):

python transcribe_voice_notes.py --contact "27833911315" --limit 5

Prerequisites

• WhatsApp MCP server running and authenticated
• gpt-4o-mini-transcribe model access
• media/audio/ folder exists in vault

Data Extraction

The skill performs these steps:

1. Download: Gets all audio messages from specified WhatsApp contact
2. Copy: Moves audio files to media/audio/ for processing
3. Transcribe: Uses gpt-4o-mini-transcribe to convert audio to text
4. Correct: Applies mining site name corrections from built-in dictionary
5. Output: Returns corrected transcriptions ready for task creation

Common Mis-Transcriptions (Corrections Applied)

Mis-Transcribed	Actual Term	Notes
Trinity	Nchwaning 3 / N3	Most common - sounds like "Trinity"
Hutt	Haut	Plant/equipment location
Ubuntu	Update	When discussing software/systems
N2/NCH2	Nchwaning 2	Sometimes transcribed as different letters
Glory/Gloria	Gloria	Mine site name (usually correct)

Example Outputs

Input: Voice note: "This is an overview of the Trinity underground plant conveyors."

Processed: "This is an overview of the Nchwaning 3 underground plant conveyors."

Input: Voice note: "Why only ten faces blasted?"

Processed: "Why only ten faces blasted?" (No corrections needed)

Output Format

Returns array of corrected transcriptions:

{
  "transcriptions": [
    {
      "timestamp": "2025-11-26 06:49",
      "original": "This is an overview of the Trinity underground plant conveyors.",
      "corrected": "This is an overview of the Nchwaning 3 underground plant conveyors.",
      "corrections_applied": ["Trinity → Nchwaning 3"],
      "file_name": "voice_20251126_065442.ogg"
    }
  ]
}

Validation Checks

• ✓ All audio files successfully downloaded
• ✓ All files copied to media/audio/
• ✓ All transcriptions generated
• ✓ Common corrections applied
• ✓ Output formatted for task creation

Common Issues & Troubleshooting

Issue: "File not found" on transcription

• Solution: Ensure file was copied to media/audio/ before transcribing

Issue: Correction not applied

• Solution: Check if term is in corrections dictionary; add new terms as needed

Issue: Wrong contact selected

• Solution: Verify phone number format (use 27XXXXXXXXX format without +)

Next Steps After Transcription

1. Review corrected transcriptions
2. Create tasks from voice note content
3. Add to task list with appropriate tags (#site, #priority)
4. Link to relevant projects if applicable

Advanced: Adding New Corrections

Edit the corrections_dict in transcribe_voice_notes.py:

corrections_dict = {
    'trinity': 'Nchwaning 3',
    'hutt': 'Haut',
    'ubuntu': 'Update',
    # Add new corrections here
}

Then restart the skill for changes to take effect.