vrcencode.py

#!/usr/bin/env python3
import ffmpeg
import sys
import fcntl
import os
infile, outfile = sys.argv[1:]

# try to work around dumb dual audio files
probe = ffmpeg.probe(infile)
audio_track = 0
audio_count = 0
audio_stream = None
sub_type = 'text'
for stream in probe['streams']:
    # try to select jpn audio over english
    if stream['codec_type'] == 'audio':
        tags = stream['tags']
        if tags and 'language' in tags and tags['language'] == 'jpn':
            audio_track = audio_count
            audio_stream = stream
        audio_count += 1
    if stream['codec_type'] == 'subtitle':
        if stream['codec_name'] == 'dvd_subtitle':
            sub_type = 'dvd'

print(f"selecting {audio_track=} {audio_stream=}", file=sys.stderr)

ffin = ffmpeg.input(infile)
ffv = ffin.video.filter('format', 'yuv420p')
# assume subtitle track matches the audio track
if sub_type == 'text':
    ffv = ffv.filter('subtitles', filename=infile, stream_index=audio_track)
else:
    # burn the dvd subs on the image
    subs = ffin[f's:{audio_track}']
    s2r = ffmpeg.filter_multi_output([subs, ffv], 'scale2ref')
    ffv = s2r[1].overlay(s2r[0])

opts = {
        'c:v': 'libx264',
        'preset': 'medium',
        'tune': 'animation',
        'movflags': 'faststart',
        'crf': '28',
        'c:a': 'aac',
        'b:a': '160k',
        'ac': '1',
        #'t': '60'
        }
audio = ffin[f'a:{audio_track}']

stream = ffmpeg.output(ffv, audio, outfile, **opts)

print(f"will run {' '.join(stream.get_args())=}", file=sys.stderr)

# basic flock queueing for multiple invocations
with open("/tmp/vrcencode", "w+") as f:
    fcntl.flock(f, fcntl.LOCK_EX)
    stream.run()