webm file supported

sentence_api/processing.py

@@ -29,6 +29,60 @@ from .transcription import (
 logger = logging.getLogger(__name__)
 MOSS_ALGORITHM_VERSION = "moss-punctuation-v3"
 SILENCE_ALGORITHM_VERSION = "silence-rms-v1"
+_CUES_ID = b"\x1e\x53\xb6\x80"
+_CLUSTER_ID = b"\x1f\x43\xb6\x75"
+_HEADER_SCAN_SIZE = 64 * 1024
+_WEBM_SUFFIXES = {".webm", ".mkv"}
+
+
+def _webm_cues_at_front(filepath: Path) -> bool:
+    """Return True if WebM/MKV CUES appear before the first Cluster."""
+    try:
+        with open(filepath, "rb") as f:
+            header = f.read(_HEADER_SCAN_SIZE)
+    except OSError:
+        return True  # can't read → assume OK, don't touch
+    cues_pos = header.find(_CUES_ID)
+    cluster_pos = header.find(_CLUSTER_ID)
+    if cues_pos != -1 and (cluster_pos == -1 or cues_pos < cluster_pos):
+        return True
+    return False
+
+
+def _ensure_webm_cues_front(filepath: Path) -> None:
+    """Re-mux a WebM/MKV file (stream copy) if CUES are not at the front.
+
+    YouTube-dl / yt-dlp writes CUES at the end by default, which slows HTTP
+    Range seeking. This runs once during upload processing and takes seconds.
+    """
+    if filepath.suffix.lower() not in _WEBM_SUFFIXES:
+        return
+    if _webm_cues_at_front(filepath):
+        return
+    ffmpeg = shutil.which("ffmpeg")
+    if not ffmpeg:
+        logger.warning("ffmpeg not found; skipping WebM CUES optimization.")
+        return
+    tmp = filepath.with_suffix(".cues_tmp" + filepath.suffix)
+    try:
+        result = subprocess.run(
+            [ffmpeg, "-y", "-loglevel", "error",
+             "-i", str(filepath), "-c", "copy",
+             "-cues_to_front", "1", str(tmp)],
+            capture_output=True, timeout=300,
+        )
+        if result.returncode == 0 and tmp.is_file() and tmp.stat().st_size > 0:
+            tmp.replace(filepath)
+            logger.info("WebM CUES moved to front: %s", filepath.name)
+        else:
+            logger.warning(
+                "WebM CUES re-mux failed for %s: rc=%d stderr=%s",
+                filepath.name, result.returncode, result.stderr.decode()[-200:],
+            )
+    except (subprocess.TimeoutExpired, OSError) as exc:
+        logger.warning("WebM CUES re-mux error for %s: %s", filepath.name, exc)
+    finally:
+        tmp.unlink(missing_ok=True)
 
 
 class VideoProcessor:
@@ -55,6 +109,7 @@ class VideoProcessor:
         self.repository.mark_processing(video_hash)
         work_path: Optional[Path] = None
         try:
+            _ensure_webm_cues_front(media_path)
             if self.transcriber.available:
                 work_path = self.settings.work_dir / f"{video_hash}-{uuid.uuid4().hex}.wav"
                 extract_audio(media_path, work_path)