def get_srt_pipe_maker(
args: argparse.Namespace, srtin: Optional[str]
) -> Callable[[Optional[float]], Union[Pipeline, Callable[[float], Pipeline]]]:
if srtin is None:
srtin_format = 'srt'
else:
srtin_format = os.path.splitext(srtin)[-1][1:]
parser = make_subtitle_parser(fmt=srtin_format, caching=True, **args.__dict__)
return lambda scale_factor: make_subtitle_speech_pipeline(
**override(args, scale_factor=scale_factor, parser=parser)
)
def make_reference_pipe(args: argparse.Namespace) -> Pipeline:
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
return Pipeline([
(
"speech_extract",
VideoSpeechTransformer(
vad=vad,
sample_rate=SAMPLE_RATE,
frame_rate=args.frame_rate,
non_speech_label=args.non_speech_label,
start_seconds=args.start_seconds,
ffmpeg_path=args.ffmpeg_path,
ref_stream=ref_stream,
vlc_mode=args.vlc_mode,
gui_mode=args.gui_mode,
),
),
])