admin-valentin/WhisperLiveKit
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

vac model is loaded in TranscriptionEngine, and by default

Browse source
This commit is contained in:
Quentin Fuxa 2025-08-17 00:34:25 +02:00
parent 0f2eba507e
commit 7fe0353260
3 changed files with 20 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
whisperlivekit/audio_processor.py
View file

@ -66,10 +66,12 @@ class AudioProcessor:
self.asr = models.asr self.asr = models.asr
self.tokenizer = models.tokenizer self.tokenizer = models.tokenizer
self.diarization = models.diarization self.diarization = models.diarization
import torch self.vac_model = models.vac_model
model, _ = torch.hub.load(repo_or_dir="snakers4/silero-vad", model="silero_vad") if self.args.vac:
self.vac = FixedVADIterator(model) self.vac = FixedVADIterator(models.vac_model)
self.vac.reset_states() else:
self.vac = None
self.ffmpeg_manager = FFmpegManager( self.ffmpeg_manager = FFmpegManager(
sample_rate=self.sample_rate, sample_rate=self.sample_rate,
channels=self.channels channels=self.channels
@ -218,11 +220,14 @@ class AudioProcessor:
# Process audio chunk # Process audio chunk
pcm_array = self.convert_pcm_to_float(self.pcm_buffer[:self.max_bytes_per_sec]) pcm_array = self.convert_pcm_to_float(self.pcm_buffer[:self.max_bytes_per_sec])
self.pcm_buffer = self.pcm_buffer[self.max_bytes_per_sec:] self.pcm_buffer = self.pcm_buffer[self.max_bytes_per_sec:]
res = self.vac(pcm_array)
res = None
end_of_audio = False end_of_audio = False
silence_buffer = None silence_buffer = None
if self.args.vac:
res = self.vac(pcm_array)
if self.silence: if self.silence:
print('NO AUDIO') print('NO AUDIO')

9
whisperlivekit/core.py
View file

@ -34,7 +34,7 @@ class TranscriptionEngine:
"lan": "auto", "lan": "auto",
"task": "transcribe", "task": "transcribe",
"backend": "faster-whisper", "backend": "faster-whisper",
"vac": False, "vac": True,
"vac_chunk_size": 0.04, "vac_chunk_size": 0.04,
"log_level": "DEBUG", "log_level": "DEBUG",
"ssl_certfile": None, "ssl_certfile": None,
@ -82,6 +82,11 @@ class TranscriptionEngine:
self.asr = None self.asr = None
self.tokenizer = None self.tokenizer = None
self.diarization = None self.diarization = None
self.vac_model = None
if self.args.vac:
import torch
self.vac_model, _ = torch.hub.load(repo_or_dir="snakers4/silero-vad", model="silero_vad")
if self.args.transcription: if self.args.transcription:
if self.args.backend == "simulstreaming": if self.args.backend == "simulstreaming":
@ -131,7 +136,7 @@ def online_factory(args, asr, tokenizer, logfile=sys.stderr):
logfile=logfile, logfile=logfile,
) )
# warmup_online(online, args.warmup_file) # warmup_online(online, args.warmup_file)
elif args.vac: elif False: #args.vac: #vac is now handled in audio_processor
online = VACOnlineASRProcessor( online = VACOnlineASRProcessor(
args.min_chunk_size, args.min_chunk_size,
asr, asr,

4
whisperlivekit/parse_args.py
View file

@ -113,8 +113,8 @@ def parse_args():
) )
parser.add_argument( parser.add_argument(
"--vac", "--vac",
action="store_true", # action="store_true",
default=False, default=True,
help="Use VAC = voice activity controller. Recommended. Requires torch.", help="Use VAC = voice activity controller. Recommended. Requires torch.",
) )
parser.add_argument( parser.add_argument(