This is a demonstration of the live transcription API offered by Deepgram, using their in-house Nova-2 model. You can use this notebook to transcribe live audio using your device's microphone input or a streaming audio from the Internet, using a websocket connection with Deepgram's API server.
Transcribing from a microphone will unfortunately only work on your local machine, and with macOS or linux as it requires portaudio.
Speaker diarization is not available for live transcription.
The radio streaming function by running the notebook on JupyterLab / Vertex AI workbench. There are some display issues on Colab unfortunately.
Running the service requires credits for Deepgram API (provided by a trial account). Please get in touch with @yaufai.chau on Slack if credits runs out or you have futher enquiries.