This code is forked and expanded from Alexander Lopatin's repository.
It calculates a delay between the audio from OBS and Waveform.
It then merges all audio from Waveform into an mkv with video from OBS.
If subtitles, bulletin PDF, and chapters are available, it will merge those in also.
If subtitles are not available, it will generate them via the Google Translate API.
If the sample rate between OBS and Waveform don't match, it will resample the Waveform audio to match.
If the video codec isn't hevc, it will convert it.
They depend on how SoX and FFmpeg were built for your OS (more details). If it didn't work with some format for you — try WAV as experiment.
Make sure these dependences are installed:
- bash (tested with 4.4.23)
- bc (tested with 1.06.95)
- ffmpeg (tested with 4.1.3)
- fftw (tested with 3.3.6_p2)
- libsndfile (tested with 1.0.28)
- python3 (tested with 3.6.10)
- sox (tested with 14.4.2)
- autosub3 (
pip3 install) (tested with 0.1.0)
On Debian some packages may need to be installed together with -dev packages (for instance fftw with fftw-dev).
Now compile it with
make -j
python3 entrypoint.py -v <video file> -a <audio dir> [-c <chapters>] [-s <subtitles>] [-p <pdfs_dir] [-o <output>]
OR
docker run -it --rm -v $(pwd):/media -w /media image_name -v obs_video.mp4 -a waveform_project/Recorded -o merged.mkv
Copyright (C) 2018—2020 Alexander Lopatin