lelo/bethaus-app
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

segment between languages

Browse Source
This commit is contained in:
lelo 2025-06-15 22:56:11 +02:00
parent a7effaec8f
commit fde07e3830
1 changed files with 13 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
transcribe_all.py
View File

@ -147,7 +147,7 @@ def process_file(file_path, model, audio_input):
# Determine which languages to transcribe # Determine which languages to transcribe
if (detected == 'ru' and 'predigt' in file_name.lower()) or \ if (detected == 'ru' and 'predigt' in file_name.lower()) or \
(detected == 'de' and 'russisch' in file_name.lower()): (detected == 'de' and 'russisch' in file_name.lower()):
langs = ['ru', 'de'] langs = ['de', 'ru']
elif detected == 'en': # songs often mis-detected as English elif detected == 'en': # songs often mis-detected as English
langs = ['de'] langs = ['de']
elif detected in ('de', 'ru'): elif detected in ('de', 'ru'):
@ -156,17 +156,12 @@ def process_file(file_path, model, audio_input):
langs = ['ru'] langs = ['ru']
# Collect segments for combined result # Collect segments for combined result
combined_segments = [] lang_collection = {}
for lang in langs: for lang in langs:
combined_segments = []
print(f"Transcribing {format_status_path(file_path)} as {lang}", end='', flush=True) print(f"Transcribing {format_status_path(file_path)} as {lang}", end='', flush=True)
result = transcribe_file(model, audio_input, lang) result = transcribe_file(model, audio_input, lang)
# Insert a synthetic segment for language header
combined_segments.append({
'start': 0,
'text': f"\n## Transcription ({lang.upper()})",
})
# Extend with actual segments # Extend with actual segments
if isinstance(result, dict) and 'segments' in result: if isinstance(result, dict) and 'segments' in result:
combined_segments.extend(result['segments']) combined_segments.extend(result['segments'])
@ -174,6 +169,7 @@ def process_file(file_path, model, audio_input):
# If result isn't dict-of-segments, wrap entire text # If result isn't dict-of-segments, wrap entire text
text = getattr(result, 'text', None) or (result.get('text') if isinstance(result, dict) else str(result)) text = getattr(result, 'text', None) or (result.get('text') if isinstance(result, dict) else str(result))
combined_segments.append({'start': 0, 'text': text}) combined_segments.append({'start': 0, 'text': text})
lang_collection[lang] = combined_segments
# Now write out markdown using the combined segments # Now write out markdown using the combined segments
file_dir = os.path.dirname(file_path) file_dir = os.path.dirname(file_path)
@ -191,12 +187,15 @@ def process_file(file_path, model, audio_input):
"" ""
] ]
previous_text = "" previous_text = ""
for segment in combined_segments: for lang, combined_segments in lang_collection.items():
start = format_timestamp(segment.get('start', 0)) md_lines.append(f"##### Transcription ({lang.upper()})")
text = segment.get('text', '').strip() md_lines.append("---")
if text and text != previous_text: for segment in combined_segments:
md_lines.append(f"`{start}` {text}") start = format_timestamp(segment.get('start', 0))
previous_text = text text = segment.get('text', '').strip()
if text and text != previous_text:
md_lines.append(f"`{start}` {text}")
previous_text = text
# Join and post-process # Join and post-process
transcript_md = "\n".join(md_lines) transcript_md = "\n".join(md_lines)