Demo of using Whisper in Python to transcribe mic audio in real time, locally.
Bit flawed because it chops up the audio into x-second blocks, so anything said in between blocks will probably be wrong or not transcribed at all.
Uses sample code from sounddevice to get audio from the mic.
Run main_continuous.py for a continuous stream of audio, or single_block.py for a single 5-second block of audio.
Requires:
- whisper (and numpy)
- sounddevice