feat: improve audio stream selection and caching

verify_sync.py

@@ -297,7 +297,8 @@ class VerifyWindow(QtWidgets.QMainWindow):
     def __init__(self, video_path: str, wav_a_path: str, wav_b_path: str,
                  sr: int = 48000, lag_sec: float = 0.0,
                  mode: str = "short", out_dir: str = "output",
-                 temp_dir: str | None = None, volume: int = 70):
+                 temp_dir: str | None = None, volume: int = 70,
+                 cache_key_a: str | None = None, cache_key_b: str | None = None):
         super().__init__()
         self.video_path = video_path
         self.wav_a_path = wav_a_path
@@ -325,9 +326,9 @@ class VerifyWindow(QtWidgets.QMainWindow):
 
         # Load audio
         print(f"Loading audio A: {wav_a_path}")
-        self.a_orig, _ = cached_load_wav(wav_a_path, target_sr=sr, _loader=load_wav)
+        self.a_orig, _ = cached_load_wav(wav_a_path, target_sr=sr, _loader=load_wav, _cache_key=cache_key_a)
         print(f"Loading audio B: {wav_b_path}")
-        self.b_orig, _ = cached_load_wav(wav_b_path, target_sr=sr, _loader=load_wav)
+        self.b_orig, _ = cached_load_wav(wav_b_path, target_sr=sr, _loader=load_wav, _cache_key=cache_key_b)
 
         # Apply lag alignment
         self.a_aligned, self.b_aligned = apply_lag(self.a_orig, self.b_orig, self.lag_samples)
@@ -688,18 +689,44 @@ def main():
             print("ERROR: synced MKV needs at least 2 audio streams (A + B)", file=sys.stderr)
             sys.exit(1)
 
+        # Determine which streams to extract from synced MKV
+        # A's streams are first, B's delayed streams follow
+        video_a = ext.get("file_a", "")
+        a_streams_in = probe_audio_streams(video_a) if os.path.exists(video_a) else []
+        n_a = len(a_streams_in)
+        sel_a = ext.get("audio_stream_a", 0) or 0
+        sel_b = ext.get("audio_stream_b", 0) or 0
+        sel_a = int(sel_a)
+        sel_b = int(sel_b)
+        # Output indices: A's selected stream stays at sel_a, B's becomes n_a + sel_b
+        out_idx_a = sel_a
+        out_idx_b = n_a + sel_b
+        print(f"A has {n_a} streams, selected [{sel_a}] -> output [{out_idx_a}]")
+        print(f"B selected [{sel_b}] -> output [{out_idx_b}]")
+
+        if out_idx_a >= len(streams) or out_idx_b >= len(streams):
+            print(f"ERROR: computed stream indices [{out_idx_a}, {out_idx_b}] out of range "
+                  f"(file has {len(streams)} streams)", file=sys.stderr)
+            sys.exit(1)
+
         temp_dir = tempfile.mkdtemp(prefix="verify_sync_")
         wav_a = os.path.join(temp_dir, "audio_a.wav")
         wav_b = os.path.join(temp_dir, "audio_b.wav")
 
-        print(f"Extracting audio stream 0 (A) to {wav_a}...")
-        if not extract_audio_from_mkv(synced, 0, wav_a, sr):
-            print("ERROR: failed to extract audio stream 0", file=sys.stderr)
+        # Stable cache keys based on source MKV + stream index (not temp file path)
+        import hashlib
+        _mkv_mtime = os.path.getmtime(synced)
+        cache_key_a = hashlib.sha256(f"{synced}|{out_idx_a}|{sr}|{_mkv_mtime:.6f}".encode()).hexdigest()[:16]
+        cache_key_b = hashlib.sha256(f"{synced}|{out_idx_b}|{sr}|{_mkv_mtime:.6f}".encode()).hexdigest()[:16]
+
+        print(f"Extracting audio stream {out_idx_a} (A) to {wav_a}...")
+        if not extract_audio_from_mkv(synced, out_idx_a, wav_a, sr):
+            print(f"ERROR: failed to extract audio stream {out_idx_a}", file=sys.stderr)
             sys.exit(1)
 
-        print(f"Extracting audio stream 1 (B) to {wav_b}...")
-        if not extract_audio_from_mkv(synced, 1, wav_b, sr):
-            print("ERROR: failed to extract audio stream 1", file=sys.stderr)
+        print(f"Extracting audio stream {out_idx_b} (B) to {wav_b}...")
+        if not extract_audio_from_mkv(synced, out_idx_b, wav_b, sr):
+            print(f"ERROR: failed to extract audio stream {out_idx_b}", file=sys.stderr)
             sys.exit(1)
 
         video_path = synced
@@ -737,6 +764,8 @@ def main():
             else:
                 print(f"WARNING: extracted video not found ({extracted_video}), using original", file=sys.stderr)
         temp_dir = None
+        cache_key_a = None
+        cache_key_b = None
 
     print(f"Mode:    {args.mode}")
     print(f"Video:   {video_path}")
@@ -750,7 +779,7 @@ def main():
 
     w = VerifyWindow(video_path, wav_a, wav_b, sr=sr, lag_sec=lag_sec,
                      mode=args.mode, out_dir=out_dir, temp_dir=temp_dir,
-                     volume=volume)
+                     volume=volume, cache_key_a=cache_key_a, cache_key_b=cache_key_b)
     w.show()
     sys.exit(app.exec())