ffsubsync

ffsubsync copied to clipboard

I read subsync/subsync.py and it seems there's too much processing in that. Maybe we should extract a Python API layer between the CLI and core code so users can choose between using either CLI or Python API.

It would be really great and make it easier to integrate in Bazarr. :-)

I agree that refactoring the entrypoint is probably a good idea. It will take a while to do so, unfortunately. Until then, a workaround is to build up a command line argument string, import ffsubsync's parser and use it to parse this, and then call ffsubsync.run(args) similar to what we do in test_integration.py

I'll look into this, thanks!

I agree! I ran into this issue today.

video = ".../somepath/video.mp4"
sub = ".../some/path/sub.srt"

unparsed_args = [video, "-i", sub, "-o", sub]

parser = ffsubsync.make_parser()
args = parser.parse_args(unparsed_args)

ffsubsync.run(args)

The output would look like this.

[08:48:36] INFO     extracting speech segments from reference    ffsubsync.py:98
                    'somemovie.m4v'...                                                 
           INFO     Checking video for subtitles      speech_transformers.py:240
                    stream...                                                   
[08:48:37] INFO     Video file appears to lack        speech_transformers.py:245
                    subtitle stream                                             
           WARNING  [Errno 2] No such file or         speech_transformers.py:251
                    directory: 'ffprobe': 'ffprobe'                             
0it [00:00, ?it/s]
Traceback (most recent call last):
  File "sync.py", line 150, in <module>
    test = ffsubsync.run(args)
  File ".../venv/lib/python3.7/site-packages/ffsubsync/ffsubsync.py", line 99, in run
    reference_pipe.fit(args.reference)
  File ".../venv/lib/python3.7/site-packages/ffsubsync/sklearn_shim.py", line 214, in fit
    self._final_estimator.fit(Xt, y, **fit_params)
  File ".../venv/lib/python3.7/site-packages/ffsubsync/speech_transformers.py", line 319, in fit
    'Unable to detect speech. Perhaps try specifying a different stream / track, or a different vad.'
ValueError: Unable to detect speech. Perhaps try specifying a different stream / track, or a different vad.

I tried calling ffsubsync using a subprocess call, executing the command with os.system(), and tried passing "--ffmpeg-path /usr/bin" (I can verify that ffprobe and ffmpeg exist in that directory but ffsubsync refuse to acknowledge it).

So, I had the same error with all of those.

NOTE: Doing it manually from the terminal works fine. Frustrating.

Has there been any work done on making this an importable python module to be used with an api?

Is there any update on this feature?

Doesn't seem like it atm

Nope, no updates I'm afraid. Work and life have both been keeping my pretty busy (in the past year, got married, bought a house, etc), and most of my OSS efforts recently have been towards ipyflow. That's not to say I haven't forgotten about this project, and I fully intend to get back into adding new features, improvements, API support, etc -- but right now I have no concrete timeline on it.

Thanks for the update man. I understand!