Seamless Branching Audio Overhang
Posted: Mon Aug 23, 2021 7:40 pm
Hello everyone!
I have the following problem with the otherwise great tool DGDemux
:
I am trying to demux Twin Peaks S01E01 (standard cut, not international cut) from "The Entire Mystery" set (https://www.blu-ray.com/movies/Twin-Pea ... ray/93181/) with the original Dolby Digital 2.0 Stereo track. Because there are two different versions of that episode on the disc, seamless branching is utilized. For the standard version, the segments are 0, 1, 2, 41.
Now the Dolby Digital tracks on this release are often strangely encoded, because for many episodes the DD audio is ca. 30 seconds longer (containing silence) than the available video. This is not a huge issue when there is no seamless branching involved, but the same applies to (some of) those M2TS files that are used within seamless branching, specifically to segment 0. That means that between segments 0 and 1, those ~30 seconds of audio must be erased to maintain sync.
Here comes the problem: DGDemux does maintain sync, but for the first ~30s of segment 1, it uses the silent audio overhang from segment 0, resulting in ~30s of silence in the demuxed audio when there should actually be dialog.
So I would propose that in such cases, DGDemux should use the audio from the segment without overhang.
I tested other demux tools as well: MakeMKV 1.16.4 has exactly the same issue, eac3to 3.34 does not, it correctly uses the audio from segment 1. However, I don't want to use the track from eac3to because it is not as nicely synced later in the track in comparison to the DGDemux output.
Here is a visual comparison of the three different outputs, where only the eac3to output contains the encircled part of the audio:
MakeMKV log:
I have the following problem with the otherwise great tool DGDemux
:I am trying to demux Twin Peaks S01E01 (standard cut, not international cut) from "The Entire Mystery" set (https://www.blu-ray.com/movies/Twin-Pea ... ray/93181/) with the original Dolby Digital 2.0 Stereo track. Because there are two different versions of that episode on the disc, seamless branching is utilized. For the standard version, the segments are 0, 1, 2, 41.
Now the Dolby Digital tracks on this release are often strangely encoded, because for many episodes the DD audio is ca. 30 seconds longer (containing silence) than the available video. This is not a huge issue when there is no seamless branching involved, but the same applies to (some of) those M2TS files that are used within seamless branching, specifically to segment 0. That means that between segments 0 and 1, those ~30 seconds of audio must be erased to maintain sync.
Here comes the problem: DGDemux does maintain sync, but for the first ~30s of segment 1, it uses the silent audio overhang from segment 0, resulting in ~30s of silence in the demuxed audio when there should actually be dialog.
So I would propose that in such cases, DGDemux should use the audio from the segment without overhang.
I tested other demux tools as well: MakeMKV 1.16.4 has exactly the same issue, eac3to 3.34 does not, it correctly uses the audio from segment 1. However, I don't want to use the track from eac3to because it is not as nicely synced later in the track in comparison to the DGDemux output.
Here is a visual comparison of the three different outputs, where only the eac3to output contains the encircled part of the audio:
Code: Select all
AV sync issue in stream 1 at 0:52:45.245 with duration of 28.098s : encountered overlapping frame, audio skew is +28.098s
AV sync issue in stream 1 at 0:53:13.344 with duration of 28.096s : 878 frame(s) dropped to reduce audio skew to +2.583ms
AV sync issue in stream 1 at 1:30:39.809 with duration of 28.041ms : encountered overlapping frame, audio skew is +30.625ms
