forked from mrq/ai-voice-cloning
and it turned out I wasn't even using the aligned segments, kmsing now that I have to *redo* my dataset again
This commit is contained in:
parent
14779a5020
commit
999878d9c6
|
@ -1035,6 +1035,12 @@ def whisper_transcribe( file, language=None ):
|
|||
model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
|
||||
result_aligned = whisperx.align(result["segments"], model_a, metadata, file, device)
|
||||
|
||||
for i in range(len(result_aligned['segments'])):
|
||||
del result_aligned['segments'][i]['word-segments']
|
||||
del result_aligned['segments'][i]['char-segments']
|
||||
|
||||
result['segments'] = result_aligned['segments']
|
||||
|
||||
return result
|
||||
|
||||
def prepare_dataset( files, outdir, language=None, skip_existings=False, progress=None ):
|
||||
|
|
Loading…
Reference in New Issue
Block a user