Blu-ray Audio Drift (Slowly losing sync/multiple .m2ts files)

[AEmovieguy]

While creating my first HD edit, I noticed that the audio track would start off perfectly synced with the video and then progressively lose sync over time. I've come to know that this is referred to as "Audio Drifting." Blu-rays with seamless branching are divided into multiple .m2ts files and have overlapping audio frames to smooth out the transition to Maximum movie modes or re-inserted scenes. If the program joining the .m2ts files back together doesn't know to properly trim these overlapping sections of audio, you end up with an audio stream about a dozen frames or so longer than your video stream, with gaps in audio at every single m2ts merge point. Basically, the audio "drifts" out of sync.

I've come to learn that TSmuxer can not properly account for these overlapping audio frames (see Edit Update below), and I was hoping someone could recommend a guide to joining together .m2ts files of Blu-rays with seamless branching.

This is putting a hold on converting a nearly completed edit to HD, and I'd prefer not to use the Dolby Digital tracks from my DVD copies of these movies.

Any help here would be hot.

EDIT Update: When using TSmuxer, I had manually joined together the multiple .m2ts files from the stream folder instead of selecting a playlist (.mpls file). Is it just that the .mpls file would have told TSmuxer how to cut the overlappng audio frames? Re-ripping the Blu-rays to find out... :-o

 

[geminigod]

I'm pretty sure I read something about this recently but can't seem to find it....

I have never personally had to deal with that yet, but here are some thoughts:

1) You can write an avs script to have virtualdub merge them into 1 master file, though I have never tried this with video & audio combined.

2) What about just importing each segment into your video editor? Then hopefully it should be a simple matter of trimming the ends to eliminate any overlap and sticking it all together. Unfortunately this would mean having to do an intermediate rendering step, I guess, if you are trying to substitute an identical reference file into an already completed edit.

 

[AEmovieguy]

Loading the .mpls file into TSmuxer helped. Instead of the audio gradually getting progressively worse until being a half second (12 frames) behind, it is only slightly inconsistent. Compared to the Dolby Digital track on the same timeline, the Lossless track (.flac) starts off maybe a tiny fraction of a frame behind, and then it becomes 0.5 of a frame ahead, 1 frame ahead, and finally 1.5 frames ahead of the Dolby DVD track (in large chunks; not gradual).

Thanks for your feedback, Geminigod. I might end up zooming in and eyeballing the adjustments against the Dolby track on the same timeline since it's so close.

With results that close, I hesitate to say that TSmuxer with a .mpls file was unsuccessful, and I tend to question if a few milliseconds went askew at the authoring house while integrating the branching feature?

 

[Captain Khajiit]

I've come to learn that TSmuxer can not properly account for these overlapping audio frames (see Edit Update below), and I was hoping someone could recommend a guide to joining together .m2ts files of Blu-rays with seamless branching.

Don't use tsmuxer to demux a BD: use eac3to. I have never encountered a sync problem, but I have demuxed seamless branching BDs before, and it has corrected the overlaps. You can use a GUI such as HD DVD/Blu-ray Stream Extractor, if you don't fancy using the command line; there's a link to that and other GUIs to be found in the eac3to link above: you need to scroll down a bit.

 

[geminigod]

HD DVD/Blu-ray Stream Extractor comes with the eac3to download. That is what I use as well and have never had a problem. It can convert to FLAC if that is what you prefer to use.

 

[AEmovieguy]

Don't use tsmuxer to demux a BD: use eac3to. I have never encountered had a sync problem, but I have demuxed seamless branching BDs before, and it has corrected the overlaps. You can use a GUI such as HD DVD/Blu-ray Stream Extractor, if you don't fancy using the command line; there's a link to that and other GUIs to be found in the eac3to link above: you need to scroll down a bit.

Thank you both. I've downloaded Clown BD which uses eac3to to correct overlap and compared the results against the Dolby DVD track and the TSmuxer track on the timeline. The eac3to demuxed track is so close at the join points to the Dolby track (we're talking a millisecond early at this point, a millisecond behind there, perfect other times) that it's hard to complain given the fact that one is Lossy and the other Lossless. But there is an OCD part of me that wishes for that millisecond to always perfectly match up, yet to a casual listener (or any listener :embarassed this will only be audible if you overlaid the tracks and heard the slightest of reverb in certain sections.

Thanks again!

 

[geminigod]

I'm not sure if you are talking about differences in the exact same format or are comparing different formats, but at the millisecond level, it would appear there can be subtle differences in reported lengths between different codecs that have nothing to do with the raw sound data. This seems to relate more to the format or how the headers self report than anything real. I just compared an identical wav, ac3, & dts file. The wav & ac3 were rendered from Vegas. The dts was rendered from the wav with a 3rd party app (more on this in a forthcoming thread). The ac3 & dts files were identical in length. The wav file was about 10ms shorter, even though the dts was rendered from the wav!

 

[AEmovieguy]

I'm not sure if you are talking about differences in the exact same format or are comparing different formats, but at the millisecond level, it would appear there can be subtle differences in reported lengths between different codecs that have nothing to do with the raw sound data. This seems to relate more to the format or how the headers self report than anything real. I just compared an identical wav, ac3, & dts file. The wav & ac3 were rendered from Vegas. The dts was rendered from the wav with a 3rd party app (more on this in a forthcoming thread). The ac3 & dts files were identical in length. The wav file was about 10ms shorter, even though the dts was rendered from the wav!

I was comparing a Lossy Dolby Digital track ripped from a separate DVD copy against a DTS-MA track from a Blu-ray copy, one above the other, on a Sony Vegas timeline. Since the DVD doesn't have the seamless branching feature that the Blu-ray has, I assumed the DVD's Dolby Digital track would be a good way to judge how well TSmuxer/eac3to pieced together the .m2ts files and corrected the audio gaps/overlap that Blu-ray requires to inaudibly transition to the .m2ts files that make up the seamless branching feature (extended/alternate scenes; Max. Movie Mode). If certain audio overlaps are left in by mistake, the DTS-MA track could lose sync with the video at any and/or all of the points in the main film where separate .m2ts files are joined together, creating an audio track of greater length than the video that cannot simply be corrected by shifting the whole track (it gets progressively more out of sync at each .m2ts join point). The DVD's Dolby Digital track does not have to account for so much overlap, beyond perhaps joining the .VOB files; thus, I assume the timing on the Dolby Digital track will more likely be more accurate to it's source file than the DTS-MA combined and demuxed with eac3to or TSmuxer.

Compared to the Dolby Digital Track's center channel on a Sony Vegas timeline:

Manually entering the .m2ts main movie files into TSmuxer resulted in a DTS-MA track that received no audio gap/overlap correction, and the track progressively lost sync with the video until, by the end of the movie, it was a full 1.5 seconds out of sync (36 frames).

Entering the .mpls main movie playlist file into TSmuxer resulted in a DTS-MA track that randomly varied by 0.5-1.5 frames out of sync but never progressively lost sync.

Entering the .mpls main movie playlist file into Eac3to resulted in a DTS-MA track that was so close to the Dolby Digital's timing throughout as to be negligible (within a fraction of a single frame's difference, ahead or behind, at any given .m2ts join point), but slightly audible as tinny reverb in certain sections of the film when overlayed with the Dolby Digital track.

 

[geminigod]

Glad to hear that eac3to got the job done!