What does this skill do?
Automatically convert audio recordings to text via Fireworks Whisper v3. Send a URL to an audio or video file and receive a full transcription, optionally with timestamps per segment and speaker recognition.
Requirements
- Plan: Growth or higher
- Integrations: none, works directly via URL
How do you activate the skill?
- Go to Skills in your dashboard
- Find Audio Transcription and click Activate
- The skill is available right away via Telegram and the web assistant
What can you do with it?
- Transcribe a meeting recording into minutes
- Write out an interview with timestamps per fragment
- Turn a podcast episode into a blog post or social content
- Process voice notes into structured text
Example commands:
- "Transcribe this conversation: [URL]"
- "Write out this interview with timestamps: [URL]"
- "Transcribe in English with speaker recognition: [URL]"
Supported formats
MP3, WAV, M4A, OGG, FLAC, WebM, up to 200 MB per file.
Models
| Model | Speed | Accuracy | Credit cost |
|---|---|---|---|
| Turbo (default) | Fast | Good | 12 cr/min |
| Large | Slower | Maximum | 20 cr/min |
Diarisation (speaker recognition) costs 40% extra credits.
Credit cost
Credits are calculated on the actual duration of the audio file, not an estimate. After transcription the exact credits are charged.
Examples:
| Duration | Model | Credits |
|---|---|---|
| 5 minutes | Turbo | 60 cr |
| 10 minutes | Turbo | 120 cr |
| 30 minutes | Large | 600 cr |
| 60 minutes | Large + diarisation | 1,680 cr |
Language detection
Without a specified language the transcription detects the language automatically. Optionally pass a language code for higher accuracy on clearly known languages (e.g. nl, en, de).
Back to Skills marketplace