Realtime Transcription Demo

This application demonstrates real-time speech transcription using multiple services:

• OpenAI gpt-4o-mini-transcribe
• OpenAI gpt-4o-transcribe
• Deepgram nova-3
• Deepgram nova-2

Setup

1. Clone the repository
2. Install dependencies:

   npm install
   

3. Create a .env file based on .env.example and add your API keys:

   OPENAI_API_KEY=your_openai_api_key_here
   DEEPGRAM_API_KEY=your_deepgram_api_key_here
   

Running the application

Use bun to run the backend server with typescript. Bun is required.

npm run start

References

OpenAI

• https://platform.openai.com/docs/guides/speech-to-text#streaming
• https://platform.openai.com/docs/guides/realtime-transcription#handling-transcriptions
• https://platform.openai.com/docs/guides/realtime?use-case=transcription#connect-with-websockets

Deepgram

• https://developers.deepgram.com/docs/live-streaming-audio
• https://developers.deepgram.com/reference/speech-to-text-api/listen-streaming
• https://developers.deepgram.com/docs/understanding-end-of-speech-detection