LLM/WhisperLiveKit
1
0
Fork 0
mirror of https://github.com/QuentinFuxa/WhisperLiveKit.git synced 2026-03-07 14:23:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add audio partial silence in chunks handling. bump to 0.2.14.post2

Browse Source
This commit is contained in:
Quentin Fuxa
2025-11-17 22:52:00 +01:00
parent 1e67bf97f0
commit bfd60b3921
2 changed files with 35 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
pyproject.toml
View File

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
[project]
name = "whisperlivekit"
version = "0.2.14"
version = "0.2.14.post2"
description = "Real-time speech-to-text with speaker diarization using Whisper"
readme = "README.md"
authors = [

42
whisperlivekit/audio_processor.py
View File

@@ -127,6 +127,7 @@ class AudioProcessor:
self.diarization_queue = asyncio.Queue() if self.args.diarization else None
self.translation_queue = asyncio.Queue() if self.args.target_language else None
self.pcm_buffer = bytearray()
self.total_pcm_samples = 0
self.transcription_task = None
self.diarization_task = None
@@ -174,6 +175,26 @@ class AudioProcessor:
self.start_silence = None
self.last_silence_dispatch_time = None
async def _enqueue_active_audio(self, pcm_chunk: np.ndarray):
if pcm_chunk is None or pcm_chunk.size == 0:
return
if not self.diarization_before_transcription and self.transcription_queue:
await self.transcription_queue.put(pcm_chunk.copy())
if self.args.diarization and self.diarization_queue:
await self.diarization_queue.put(pcm_chunk.copy())
self.silence_duration = 0.0
def _slice_before_silence(self, pcm_array, chunk_sample_start, silence_sample):
if silence_sample is None:
return None
relative_index = int(silence_sample - chunk_sample_start)
if relative_index <= 0:
return None
split_index = min(relative_index, len(pcm_array))
if split_index <= 0:
return None
return pcm_array[:split_index]
def convert_pcm_to_float(self, pcm_buffer):
"""Convert PCM buffer in s16le format to normalized NumPy array."""
return np.frombuffer(pcm_buffer, dtype=np.int16).astype(np.float32) / 32768.0
@@ -669,25 +690,30 @@ class AudioProcessor:
pcm_array = self.convert_pcm_to_float(self.pcm_buffer[:aligned_chunk_size])
self.pcm_buffer = self.pcm_buffer[aligned_chunk_size:]
num_samples = len(pcm_array)
chunk_sample_start = self.total_pcm_samples
chunk_sample_end = chunk_sample_start + num_samples
res = None
if self.args.vac:
res = self.vac(pcm_array)
if res is not None:
if res.get("end", 0) > res.get("start", 0) and not self.silence:
silence_detected = res.get("end", 0) > res.get("start", 0)
if silence_detected and not self.silence:
pre_silence_chunk = self._slice_before_silence(
pcm_array, chunk_sample_start, res.get("end")
)
if pre_silence_chunk is not None and pre_silence_chunk.size > 0:
await self._enqueue_active_audio(pre_silence_chunk)
await self._begin_silence()
elif self.silence:
await self._end_silence()
if not self.silence:
if not self.diarization_before_transcription and self.transcription_queue:
await self.transcription_queue.put(pcm_array.copy())
await self._enqueue_active_audio(pcm_array)
if self.args.diarization and self.diarization_queue:
await self.diarization_queue.put(pcm_array.copy())
self.silence_duration = 0.0
self.total_pcm_samples = chunk_sample_end
if not self.args.transcription and not self.args.diarization:
await asyncio.sleep(0.1)