Malina Claude Opus 5 (1M context) commited on
Commit
3b71b6c
·
1 Parent(s): ac0c9e0

Film bez sciezki dzwiekowej przestaje wywracac cala analize

Browse files

Kontener bez audio wysadzal odczyt: PyAV rzuca IndexError w srodku
faster_whisper, kiedy prosi o strumien audio numer 0 w pliku, ktory go nie ma.
Wyjatek szedl do slownika i ginal, a uzytkownik dostawal "Error reading video."
w jedna sekunde. Zmierzone na zwyklym H.264 bez audio: kazda analiza takiego
pliku konczyla sie tak samo, czyli nagranie ekranu bez narracji, najbardziej
naturalny przypadek uzycia dla tego narzedzia, nie dzialalo w ogole.

Brak dzwieku to teraz cisza, a nie blad: load_audio zwraca pusta fale, Whisper
nie jest w ogole wolany, a synteza idzie sciezka wizualna, ktora ten kod od
poczatku deklarowal ("No speech is fine").

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Files changed (2) hide show
  1. audio.py +17 -2
  2. pipeline.py +4 -1
audio.py CHANGED
@@ -14,8 +14,23 @@ from utils import fmt_time
14
 
15
 
16
  def load_audio(media_path: str) -> np.ndarray:
17
- """Decode any container to float32 mono 16 kHz waveform (PyAV, no temp file)."""
18
- return decode_audio(media_path, sampling_rate=SAMPLE_RATE)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
 
20
 
21
  def media_duration(media_path: str) -> float:
 
14
 
15
 
16
  def load_audio(media_path: str) -> np.ndarray:
17
+ """Decode any container to float32 mono 16 kHz waveform (PyAV, no temp file).
18
+
19
+ A file with NO audio stream comes back as silence rather than an exception. PyAV
20
+ raises IndexError deep inside faster_whisper when it asks for audio stream 0 of a
21
+ container that has none, and that exception used to take the whole ingest down:
22
+ every silent clip answered "Error reading video." in about a second, with the
23
+ reason swallowed. Measured on a plain H.264 file with no audio track, 2026-09-21.
24
+
25
+ Silence is the honest input here. The pipeline already handles "no speech" and
26
+ falls back to visual-only synthesis, which is exactly what a screen recording
27
+ without narration needs."""
28
+ try:
29
+ return decode_audio(media_path, sampling_rate=SAMPLE_RATE)
30
+ except (IndexError, StopIteration):
31
+ print(f"[NINA] no audio stream in {str(media_path).split('/')[-1]}, "
32
+ f"continuing from the picture alone", flush=True)
33
+ return np.zeros(0, dtype=np.float32)
34
 
35
 
36
  def media_duration(media_path: str) -> float:
pipeline.py CHANGED
@@ -100,7 +100,10 @@ def ingest_video(video_path: str, stage: "_Stage | None" = None) -> dict:
100
  # WHISPER fast pass, then a fallback ladder when it looks unreliable:
101
  # 1. a clean embedded subtitle track (best, if the file has one), else
102
  # 2. vocal isolation + re-transcription (pull the voice out of the music).
103
- segments = transcribe_fast(audio)
 
 
 
104
  if segments:
105
  rel = transcript_reliability(segments)
106
  if not rel["reliable"]:
 
100
  # WHISPER fast pass, then a fallback ladder when it looks unreliable:
101
  # 1. a clean embedded subtitle track (best, if the file has one), else
102
  # 2. vocal isolation + re-transcription (pull the voice out of the music).
103
+ # No waveform at all means the container had no audio stream. Whisper has
104
+ # nothing to do with an empty array, and asking it anyway is how a silent
105
+ # screen recording turns into a failed build.
106
+ segments = transcribe_fast(audio) if getattr(audio, "size", 0) else []
107
  if segments:
108
  rel = transcript_reliability(segments)
109
  if not rel["reliable"]: