0x2620/pandora_transcribe
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

support multipart items

Browse source
This commit is contained in:
j 2025-01-25 14:36:15 +05:30
parent f28d37e33e
commit eb2d12a905
1 changed files with 15 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

22
transcribe.py
View file

@ -150,15 +150,23 @@ def extract_subtitles(item, user, layer, translate, gpu=False, join_sentences=Fa
if not item.streams():
logger.error("skip item without media %s: %s", item.public_id)
return False
src = item.streams()[0].media.path
response = run_whisper(src, language, translate, gpu, model)
if not response:
logger.error("extract failed for %s", item.public_id)
return False
annotations = prepare_annotations(response, join_sentences=join_sentences)
offset = 0
annotations = []
for stream in item.streams():
src = stream.media.path
response = run_whisper(src, language, translate, gpu, model)
if not response:
logger.error("extract failed for %s", item.public_id)
return False
for annotation in prepare_annotations(response, join_sentences=join_sentences):
if offset:
annotation['in'] += offset
annotation['out'] += offset
annotations.append(annotation)
offset += stream.duration
if not annotations:
return False
if language and language != "en":
for annotation in annotations:
annotation["value"] = '<span lang="%s">%s</span>' % (