Speech to Text Transcriber

Upload an audio file and receive a transcript using Workers AI Whisper (@cf/openai/whisper-large-v3-turbo). Validates file size and content type before calling the model.

Features

• POST /transcribe - Multipart upload with field audio
• Whisper model - @cf/openai/whisper-large-v3-turbo
• Validation - Max 2 MB, audio/* types only
• Timing - Returns durationMs for the transcription call

API Reference

POST /transcribe

Transcribe an uploaded audio file.

audio file (required, multipart)

Audio file (max 2 MB). Accepted types include audio/mpeg, audio/wav, audio/webm, audio/ogg, and other audio/* MIME types.

Example Request

curl -X POST "https://your-worker.workers.dev/transcribe" \
  -F "audio=@sample.mp3"

Success Response

{
  "text": "Hello, this is a test recording.",
  "language": "en",
  "durationMs": 842
}

Error Codes

• 400 - Missing or invalid audio (INVALID_AUDIO)
• 502 - Transcription failed or empty result (TRANSCRIBE_ERROR)

Use Cases

• Add speech-to-text to edge apps without external API keys
• Prototype voice note or meeting transcription workflows
• Learn Workers AI audio model integration patterns

Limitations

• Max upload size 2 MB per request (Whisper model limits apply)
• No chunking for long recordings; split large files client-side
• Requires Workers AI enabled on your Cloudflare account

Deployment

curl -X POST "https://your-worker.workers.dev/transcribe" -F "audio=@sample.mp3"

Local Development

cd apps/experiments/speech-to-text-transcriber
npm install
npm run dev

Configuration

Cloudflare Features Used

• Workers - Edge compute runtime
• Workers AI - Whisper speech-to-text