hello-transcription

Real-time speech transcription using OpenAI's Realtime API - a demonstration of transcription-only mode without AI responses.

• Real-time Transcription: Live speech-to-text conversion as you speak
• Multiple Models: Choose between GPT-4o Transcribe, GPT-4o Mini Transcribe, or Whisper-1
• Language Support: Transcribe in multiple languages (English, Spanish, French, German, Italian, Portuguese, Chinese, Japanese, Korean)
• Voice Activity Detection (VAD): Automatic detection of speech segments
• Logprobs Support: Optional confidence scores for transcriptions
• Split View: See transcriptions and event logs side-by-side

This app uses OpenAI's Realtime API in transcription-only mode:

1. Your voice is captured via WebRTC
2. Audio is streamed to OpenAI's transcription service
3. Transcriptions are returned in real-time
4. No AI responses are generated (transcription only)

• GPT-4o Transcribe: Streaming transcription with incremental updates
• GPT-4o Mini Transcribe: Smaller model, streaming transcription
• Whisper-1: Complete transcription after each speech segment (no streaming)

• Model: Select the transcription model
• Language: Choose the primary language for better accuracy
• VAD: Enable/disable automatic voice activity detection
• Logprobs: Include confidence scores (for advanced use)

1. Click "Start" to begin transcription
2. Allow microphone access when prompted
3. Start speaking - transcriptions appear in real-time
4. Partial transcriptions update as you speak
5. Final transcriptions are marked in green
6. Click "Stop" to end the session

• GET / - Serves the transcription interface
• POST /rtc - Creates a WebRTC transcription session
• POST /observer/:callId - WebSocket observer for transcription events

Set in your Val Town environment:

• OPENAI_API_KEY - Your OpenAI API key (required)

# Install Deno
curl -fsSL https://deno.land/install.sh | sh

# Run locally
deno run --allow-all main.tsx

1. Fork/remix this val on Val Town
2. Add your OPENAI_API_KEY to Val Town secrets
3. Your app will be available at https://[your-val-name].val.run

The app uses OpenAI's Realtime API in transcription mode:

• Session type: transcription (not realtime)
• Audio format: PCM16
• Noise reduction: Near-field (optimized for close microphones)
• WebRTC data channel for receiving transcription events

1. input_audio_buffer.committed - Audio chunk received
2. conversation.item.input_audio_transcription.delta - Partial transcription
3. conversation.item.input_audio_transcription.completed - Final transcription

Built with OpenAI's Realtime API for transcription-only use cases.