Repository files navigation

A real-time streaming transcription demo using Deepgram's Flux model. This demo streams live audio from BBC World Service and provides real-time transcription with word-level confidence scoring and color-coded output.

🔑 To access the Deepgram API you will need afree Deepgram API Key.

Before running this demo, make sure you have:

• Python 3.8+ installed on your system
• ffmpeg installed for audio conversion
  • macOS:brew install ffmpeg
  • Ubuntu/Debian:sudo apt update && sudo apt install ffmpeg
  • Windows: Download fromffmpeg.org or usewinget install ffmpeg

1. Clone the repository:

   git clone https://github.com/deepgram/deepgram-demos-flux-streaming-transcription.gitcd deepgram-demos-flux-streaming-transcription

2. Install Python dependencies:

   pip install deepgram-sdk python-dotenv aiohttp ffmpeg-python

3. Set up your environment variables:

   Create a.env file in the project root:

   echo"DEEPGRAM_API_KEY=your_deepgram_api_key_here"> .env

   Replaceyour_deepgram_api_key_here with your actual Deepgram API key.

Start the streaming transcription:

python main.py

The demo will:

1. Connect to Deepgram's Flux streaming service
2. Stream live audio from BBC World Service
3. Convert the audio stream to the correct format using ffmpeg
4. Display real-time transcription with:
   • 🎤 Main transcript text
   • 📝 Word-level confidence scores with color coding:

To stop the demo, pressCtrl+C.

You can learn more about Deepgram APIs atdevelopers.deepgram.com.

We love to hear from you! If you have questions:

• Join the Deepgram Github Discussions Community
• Join the Deepgram Discord Community