chrissewell/whisperX
1
0
Fork 0
mirror of https://github.com/m-bain/whisperX.git synced 2025-07-01 18:17:27 -04:00
Code Issues Packages Projects Releases Wiki Activity

make diarization faster

Browse Source
This commit is contained in:
Dudu Asulin
2023-08-02 10:11:43 +03:00
committed by GitHub
parent d80b98601b
commit 8de0e2af51
3 changed files with 10 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@ -177,7 +177,7 @@ print(result["segments"]) # after alignment
diarize_model = whisperx.DiarizationPipeline(use_auth_token=YOUR_HF_TOKEN, device=device)
# add min/max number of speakers if known
diarize_segments = diarize_model(audio_file)
diarize_segments = diarize_model(audio)
# diarize_model(audio_file, min_speakers=min_speakers, max_speakers=max_speakers)
result = whisperx.assign_word_speakers(diarize_segments, result)

8
whisperx/diarize.py
View File

@ -4,6 +4,8 @@ from pyannote.audio import Pipeline
from typing import Optional, Union
import torch
from .audio import SAMPLE_RATE
class DiarizationPipeline:
def __init__(
self,
@ -16,7 +18,11 @@ class DiarizationPipeline:
self.model = Pipeline.from_pretrained(model_name, use_auth_token=use_auth_token).to(device)
def __call__(self, audio, min_speakers=None, max_speakers=None):
segments = self.model(audio, min_speakers=min_speakers, max_speakers=max_speakers)
audio_data = {
'waveform': torch.from_numpy(audio[None, :]),
'sample_rate': SAMPLE_RATE
}
segments = self.model(audio_data, min_speakers=min_speakers, max_speakers=max_speakers)
diarize_df = pd.DataFrame(segments.itertracks(yield_label=True))
diarize_df['start'] = diarize_df[0].apply(lambda x: x.start)
diarize_df['end'] = diarize_df[0].apply(lambda x: x.end)

3
whisperx/transcribe.py
View File

@ -202,7 +202,8 @@ def cli():
results = []
diarize_model = DiarizationPipeline(use_auth_token=hf_token, device=device)
for result, input_audio_path in tmp_results:
diarize_segments = diarize_model(input_audio_path, min_speakers=min_speakers, max_speakers=max_speakers)
audio = load_audio(input_audio_path)
diarize_segments = diarize_model(audio, min_speakers=min_speakers, max_speakers=max_speakers)
result = assign_word_speakers(diarize_segments, result)
results.append((result, input_audio_path))
# >> Write