Skip to main content

Announcing speech-to-text: CLI Audio Transcription Using Gemini

· 2 min read
Gareth Andrew
Gareth Andrew
Chief Auto Coder

I've just released @autocode2/speech-to-text, a Node.js library and CLI tool that makes it easy to transcribe speech using Google's Gemini API.

Gemini is extremely cheap (or free) for audio processing, 1 minute of audio is 1,500 tokens, or 90k tokens/hour. Google's free tier (if you don't mind sharing your audio with Google) offers:

  • 1.5 Flash: 1 million free tokens per minute
  • 1.5 Pro: 32k free tokens per minute (though only 50 requests per day)

Either option allows for a lot of transcription for free, and the quality of Flash is quite reasonable (my accent is hard to parse for many speech-to-text systems, but Flash does a good job).

If you'd rather pay for your tokens (and keep your audio private):

  • 1.5 Flash: $0.075 per million tokens, or 0.01125 cents per minute, or less than 1 cent per hour
  • 1.5 Pro: $1.25 per million tokens, or 0.1875 cents per minute, or 11.25 cents per hour

Quick Start​

The quickest way to try it out is with npx:

npx @autocode2/speech-to-text --api-key YOUR_API_KEY

This will record from your microphone until you press Enter, then transcribe the audio using Gemini's flash model.

Features​

  • Simple CLI Interface: Record and transcribe with a single command
  • File Support: Transcribe existing audio files or save recordings
  • Flexible Output: Text or JSON output, perfect for scripting
  • Library Integration: Easy to integrate into Node.js projects
  • Custom Prompts: Fine-tune transcription behavior

Use Cases​

You can use it for:

  • Quick voice notes
  • Transcribing meetings or lectures
  • Processing audio files in batch
  • Building transcription into your own tools
  • Experimenting with Gemini's audio capabilities

Technical Details​

The tool uses:

  • sox for high-quality audio recording
  • Gemini's flash model for fast, cost-effective transcription
  • Node.js streams for efficient processing

Getting Started​

Check out the GitHub repository for:

  • Full installation instructions
  • Detailed API documentation
  • Usage examples
  • Configuration options

Future Plans​

This is just the beginning - next I want to add realtime transcription.

Try it out and let me know what you think! Feedback and contributions welcome.