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

fix: speaker embedding bug (#1178)

Browse Source 
* fix: improve handling of speaker embeddings in transcribe_task

* chore: bump version to 3.4.1
This commit is contained in:
Barabazs
2025-06-25 13:55:20 +02:00
committed by GitHub
parent b93e9b6f57
commit ffedc5cdf0
3 changed files with 10 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
pyproject.toml
View File

@ -2,7 +2,7 @@
urls = { repository = "https://github.com/m-bain/whisperx" } urls = { repository = "https://github.com/m-bain/whisperx" }
authors = [{ name = "Max Bain" }] authors = [{ name = "Max Bain" }]
name = "whisperx" name = "whisperx"
version = "3.4.0" version = "3.4.1"
description = "Time-Accurate Automatic Speech Recognition using Whisper." description = "Time-Accurate Automatic Speech Recognition using Whisper."
readme = "README.md" readme = "README.md"
requires-python = ">=3.9, <3.13" requires-python = ">=3.9, <3.13"

2
uv.lock generated
View File

@ -2788,7 +2788,7 @@ wheels = [
[[package]] [[package]]
name = "whisperx" name = "whisperx"
version = "3.4.0" version = "3.4.1"
source = { editable = "." } source = { editable = "." }
dependencies = [ dependencies = [
{ name = "ctranslate2" }, { name = "ctranslate2" },

9
whisperx/transcribe.py
View File

@ -213,12 +213,19 @@ def transcribe_task(args: dict, parser: argparse.ArgumentParser):
results = [] results = []
diarize_model = DiarizationPipeline(model_name=diarize_model_name, use_auth_token=hf_token, device=device) diarize_model = DiarizationPipeline(model_name=diarize_model_name, use_auth_token=hf_token, device=device)
for result, input_audio_path in tmp_results: for result, input_audio_path in tmp_results:
diarize_segments, speaker_embeddings = diarize_model( diarize_result = diarize_model(
input_audio_path, input_audio_path,
min_speakers=min_speakers, min_speakers=min_speakers,
max_speakers=max_speakers, max_speakers=max_speakers,
return_embeddings=return_speaker_embeddings return_embeddings=return_speaker_embeddings
) )
if return_speaker_embeddings:
diarize_segments, speaker_embeddings = diarize_result
else:
diarize_segments = diarize_result
speaker_embeddings = None
result = assign_word_speakers(diarize_segments, result, speaker_embeddings) result = assign_word_speakers(diarize_segments, result, speaker_embeddings)
results.append((result, input_audio_path)) results.append((result, input_audio_path))
# >> Write # >> Write