# transcribe

> Transcribe audio files using Deepgram API. Use when user wants to transcribe audio/video files to text.

- Author: David Lin
- Repository: abundance-company/claude-code
- Version: 20260122165217
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/abundance-company/claude-code
- Web: https://mule.run/skillshub/@@abundance-company/claude-code~transcribe:20260122165217

---

---
name: transcribe
description: Transcribe audio files using Deepgram API. Use when user wants to transcribe audio/video files to text.
user-invocable: true
disable-model-invocation: false
argument-hint: "<file_path>"
allowed-tools: Bash, Read, Write
---

# Transcribe Audio with Deepgram

Transcribe audio or video files to text using Deepgram's Nova-3 model.

## Usage

```
/transcribe <file_path>
/transcribe /path/to/audio.mp3
/transcribe ~/Downloads/meeting.m4a
```

## Supported Formats

- Audio: mp3, wav, m4a, flac, ogg, webm, aac
- Video: mp4, mov, mkv, avi (extracts audio)

## When User Requests Transcription

1. **Verify the file exists**
   ```bash
   test -f "<file_path>" && echo "✓ File exists" || echo "✗ File not found"
   ```

2. **Check for Deepgram API key**
   The key should be in `~/.claude/.env` as `DEEPGRAM_API_KEY`.
   ```bash
   source ~/.claude/.env && test -n "$DEEPGRAM_API_KEY" && echo "✓ API key found" || echo "✗ API key missing"
   ```

   If missing, ask user to add it to `~/.claude/.env`:
   ```
   DEEPGRAM_API_KEY=your_key_here
   ```
   Get a key at: https://console.deepgram.com/

3. **Run transcription**
   ```bash
   source ~/.claude/.env && python3 ~/.claude/skills/transcribe/transcribe.py "<file_path>"
   ```

4. **Display results**
   - Show the transcript text
   - Report duration and processing time
   - Note the output file location (same directory as input, with .txt extension)

## Output

- Transcript saved to: `<original_filename>.transcript.txt`
- Console shows: transcript text, duration, confidence score

## Options

The script supports these environment variables:
- `DEEPGRAM_MODEL`: Model to use (default: nova-2)
- `DEEPGRAM_LANGUAGE`: Language code (default: en, auto-detects)
- `DEEPGRAM_DIARIZE`: Enable speaker diarization (default: false)
- `DEEPGRAM_PUNCTUATE`: Add punctuation (default: true)
- `DEEPGRAM_PARAGRAPHS`: Add paragraph breaks (default: true)

Example with options:
```bash
DEEPGRAM_DIARIZE=true source ~/.claude/.env && python3 ~/.claude/skills/transcribe/transcribe.py meeting.mp3
```

## Troubleshooting

- **API key error**: Ensure `DEEPGRAM_API_KEY` is set in `~/.claude/.env`
- **File not found**: Check the file path is correct and accessible
- **Unsupported format**: Convert to mp3/wav first using ffmpeg
- **Large files**: Files over 2GB may need to be split

## Pricing

Deepgram Nova-2: ~$0.0043/minute ($0.26/hour)
See: https://deepgram.com/pricing