A stupid little helper utility for automatically and losslessly splitting and splicing together portions of audio from multiple sources. Designed primarily for syncing different versions of audio to a common video source.
Requires Python 2.7, click, and mkvmerge from MKVToolNix. Also probably requires familiarity with AviSynth and AvsPmod if you want accurate synchronization.
Your audio tracks must use the same codec, sampling rate, and channel configuration. If not, transcoding is required. For some codecs (e.g. AC-3), the bitrate should be the same as well, or you'll encounter decoding errors in some players (such as VLC).
Your videos must have the same constant frame rate when syncing with AviSynth. If they are different or one of them is variable, you must use ChangeFPS for CFR video or your source filter's frame rate arguments (e.g. fpsnum/fpsden in FFVideoSource) for VFR video to achieve a common frame rate.
Let's say we have two different versions of a media file—a Japanese version and an English version. We want to sync the audio from the English version to the Japanese video, but the Japanese version contains a few extra/different parts. To solve this, we'll create a new English audio track that takes these parts from the Japanese audio.
To find out which parts we need and from where, we write an AviSynth script that effectively syncs the English video to the Japanese video, taking parts from the Japanese video where needed.
eng = FFVideoSource("foo_eng.mkv").FFInfo()
jpn = FFVideoSource("foo_jpn.mkv").FFInfo()
cuts = eng.Trim(3, 1200) \
+ jpn.Trim(1198, 3378) \
+ eng.Trim(1898, 30761) \
+ jpn.Trim(32243, 34401) \
+ eng.Trim(32370, 32728)
StackHorizontal(cuts, jpn)
After that's done, we demux the audio tracks, transcode them if necessary, and convert the trims to a manifest file.
foo_{}.mka
eng 3 1200
jpn 1198 3378
eng 1898 30761
jpn 32243 34401
eng 32370 32728
The first line is a "template" filename. For each line, multisplice will take the template and replace the {} with the first part of the line (e.g. eng to foo_eng.mka). The second part of each line is the trim points, measured in video frames (inclusive).
Then we let multisplice (actually mostly mkvmerge) do its magic:
# ./multisplice.py -f 24000/1001 manifest.txt
The -f parameter specifies the frame rate at which to convert the trim points into timestamps. Since our videos are NTSC film (approximately 23.976 frames per second), we use 24000/1001.
By default, the output is saved to a filename based on the template, with the {} replaced with "spliced" (in this case, the output filename is foo_spliced.mka). You can also use the -o parameter to specify a filename.