2021-11-22, 03:22 AM
I was trying to make a hybrid release for the movie "Source Code" (2011):
- Video: US 4K Blu-ray
- Audio: DTS-HD MA 5.1 from US Blu-ray release (the initial one from 2011)
in order to get away with the revisionist Dolby Atmos upmix from the 4K Blu-ray.*
The Blu-ray is "protected" with playlist obfuscation so that it employs seamless branching of several scenes, so I used DGDemux for stitching together the DTS-HD MA track as it is always under ~6ms of desync (better than e.g. MakeMKV).
In theory, the audio swap should be simple, because spot-checking the video tracks in VirtualDub shows that they are in-sync (frame-exact) up to the closing credits. Also, they both run with 23.976fps (per MediaInfo, and also VirtualDub displays exact same timecodes for a given frame-number).
As a (theoretically) redundant check, I wanted to make sure that the audio tracks are in sync, too. So I loaded both tracks into Audacity and compared the waveforms of both Center and LFE tracks. To do that, I needed to convert the THD track from the 4K to w64 (PCM) with eac3to, as Audacity does not import .thd files.
Here I noticed something strange: Not only are they out-of-sync against each other by about 40ms, but the desync even increases until the end of the movie until it reaches about 150ms of desync.
Has anybody an idea why this is happening? Are the tracks really that out of sync? I can't imagine this to be the case, as 150ms desync should be very noticeable. Or is there anything in my toolchain (mainly Audacity or eac3to) that would introduce any sync/timecode variances? Or is it possible that both versions do not actually run exactly at 23.976....fps but there are rounding errors that introduce timing variability? This really baffles me...
* I know that I can downmix the Atmos track, and in fact I have found that that is probably also the better option here because the Atmos track has better dynamic, but that is not the point here.
- Video: US 4K Blu-ray
- Audio: DTS-HD MA 5.1 from US Blu-ray release (the initial one from 2011)
in order to get away with the revisionist Dolby Atmos upmix from the 4K Blu-ray.*
The Blu-ray is "protected" with playlist obfuscation so that it employs seamless branching of several scenes, so I used DGDemux for stitching together the DTS-HD MA track as it is always under ~6ms of desync (better than e.g. MakeMKV).
In theory, the audio swap should be simple, because spot-checking the video tracks in VirtualDub shows that they are in-sync (frame-exact) up to the closing credits. Also, they both run with 23.976fps (per MediaInfo, and also VirtualDub displays exact same timecodes for a given frame-number).
As a (theoretically) redundant check, I wanted to make sure that the audio tracks are in sync, too. So I loaded both tracks into Audacity and compared the waveforms of both Center and LFE tracks. To do that, I needed to convert the THD track from the 4K to w64 (PCM) with eac3to, as Audacity does not import .thd files.
Here I noticed something strange: Not only are they out-of-sync against each other by about 40ms, but the desync even increases until the end of the movie until it reaches about 150ms of desync.
Has anybody an idea why this is happening? Are the tracks really that out of sync? I can't imagine this to be the case, as 150ms desync should be very noticeable. Or is there anything in my toolchain (mainly Audacity or eac3to) that would introduce any sync/timecode variances? Or is it possible that both versions do not actually run exactly at 23.976....fps but there are rounding errors that introduce timing variability? This really baffles me...
* I know that I can downmix the Atmos track, and in fact I have found that that is probably also the better option here because the Atmos track has better dynamic, but that is not the point here.