Whisper VAD

Whisper.cpp Speech-to-Text engine combined with Silero Voice Activity Detector. This improves transcription speed and quality, and can avoid hallucination of the model.

Run whisper_vad.py directly for transcribing any video/audio files into SRT subtitles, or import it as a library.

• ffmpeg (command)
• openblas (system library)
• cffi
• torch
• scipy
• zhconv: Chinese postprocess

1. pip install -r requirements.txt
2. make

1. git submodule update --init --recursive
2. cd whisper.cpp
3. Compile whisper.cpp to match your device
   1. cmake -B build (add any build options)
   2. cmake --build build --config Release -j8
4. pip install -r requirements.txt
5. make

python3 whisper_vad.py --help to see usage.