Bug
faster_whisper_backend.py line 79 sets:
self.vad_parameters = vad_parameters or {"threshold": 0.5}
But faster-whisper==1.1.0 (which is pinned in setup.py) renamed VadOptions.threshold to VadOptions.onset. This causes every transcription attempt to fail with:
VadOptions.__init__() got an unexpected keyword argument 'threshold'
The error repeats on every audio chunk, so no transcription output is ever produced.
Fix
Change "threshold" to "onset":
self.vad_parameters = vad_parameters or {"onset": 0.5}
Steps to reproduce
pip install whisper-live # installs faster-whisper==1.1.0
python3 run_server.py --port 9090 --backend faster_whisper
# connect any client — server logs the error on every audio chunk