Function make_reference_pipe

ffsubsync/ffsubsync.py:303–400

(
    args: argparse.Namespace,
    progress_handler: Optional[Callable[[ProgressInfo], None]] = None,
)

def make_reference_pipe(
    args: argparse.Namespace,
    progress_handler: Optional[Callable[[ProgressInfo], None]] = None,
) -> Pipeline:
    pgs_stream = getattr(args, "pgs_ref_stream", None)
    if pgs_stream is not None:
        # "auto" (bare --pgs-ref-stream flag) → let PGSSpeechTransformer auto-detect
        resolved_stream: Optional[str] = None if pgs_stream == "auto" else pgs_stream
        if resolved_stream is not None and not resolved_stream.startswith("0:"):
            resolved_stream = "0:" + resolved_stream
        return Pipeline(
            [
                (
                    "speech_extract",
                    PGSSpeechTransformer(
                        sample_rate=SAMPLE_RATE,
                        start_seconds=args.start_seconds,
                        ffmpeg_path=args.ffmpeg_path,
                        ref_stream=resolved_stream,
                        gui_mode=args.gui_mode,
                    ),
                ),
            ]
        )
    ref_format = _ref_format(args.reference)
    if ref_format in SUBTITLE_EXTENSIONS:
        if args.vad is not None:
            logger.warning("Vad specified, but reference was not a movie")
        return cast(
            Pipeline,
            make_subtitle_speech_pipeline(
                fmt=ref_format,
                **override(args, encoding=args.reference_encoding or DEFAULT_ENCODING),
            ),
        )
    elif ref_format in ("npy", "npz"):
        if args.vad is not None:
            logger.warning("Vad specified, but reference was not a movie")
        return Pipeline(
            [("deserialize", DeserializeSpeechTransformer(args.non_speech_label))]
        )
    else:
        vad = args.vad or DEFAULT_VAD
        if args.reference_encoding is not None:
            logger.warning(
                "Reference srt encoding specified, but reference was a video file"
            )
        ref_stream = args.reference_stream
        if ref_stream is not None and not ref_stream.startswith("0:"):
            ref_stream = "0:" + ref_stream
        if getattr(args, "multi_segment_sync", False):
            return Pipeline(
                [
                    (
                        "speech_extract",
                        MultiSegmentVideoSpeechTransformer(
                            vad=vad,
                            sample_rate=SAMPLE_RATE,