VitekDev
/
ai-voice-cloning
forked from mrq/ai-voice-cloning
1
0
Fork 0
Code Pull Requests Activity

fixes for whisperx batching

master
mrq 2023-03-22 19:53:42 +07:00
parent 4056a27bcb
commit 8877960062
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/utils.py
Unescape Escape View File

@ -1168,7 +1168,7 @@ def whisper_transcribe( file, language=None ):
device = "cuda" if get_device_name() == "cuda" else "cpu"
if whisper_vad:
if args.whisper_batchsize > 1:
result = whisperx.transcribe_with_vad_parallel(whisper_model, file, whisper_vad, batch_size=args.whisper_batchsize)
result = whisperx.transcribe_with_vad_parallel(whisper_model, file, whisper_vad, batch_size=args.whisper_batchsize, language=language, task="transcribe")
else:
result = whisperx.transcribe_with_vad(whisper_model, file, whisper_vad)
else:
@ -1192,6 +1192,10 @@ def whisper_transcribe( file, language=None ):
del result_aligned['segments'][i]['char-segments']
result['segments'] = result_aligned['segments']
result['text'] = []
for segment in result['segments']:
result['text'].append(segment['text'].strip())
result['text'] = " ".join(result['text'])
return result