DBZ audio sync issue, remastered sets. different delays NTSC DVD

[xekon]

I see what your talking about when I checked episode 65. The problem with 65 is not that it is more out of sync(well it is but not for the reason you might think), the actual issue is that there was a transfer error on the Japanese version where there were about 20-30 extra duplicate frames, when Funimation did the english version they cut these frames out, so If you are using the Funimation english track with the R2 DragonBox video, the solution I am going to use is 3 extra trims to cut the frames from the R2 Dragonbox Video that are not present in the Funimation english video.

I do think there are some episodes that can look a lot more natural and play better if the delay is tweeked some, you basically just have to pick what is most important though, sound effects lining up or the voices.

[Corpsecreate]

They aren't duplicate frames. The scene of gohan trying to get up goes longer on the Dragon Box than it does on the orange bricks, no frames are duplicates. For some reason, Funimation removed some of this scene (my guess is because this particular scene has one frame of interlaced footage).

My solution to this problem is not the same as yours because I am also adding the japanese audio to my files which will cause that to desync (unless I cut it). Instead, I loop the english audio for 17 frames, the particular loop point I used makes it hard to notice that it was ever edited.

[xekon]

Ahh thats what it was, and that's a pretty good idea for that!

[xekon]

OK, so I figured something else out today. The points at which the audio goes out of alignment on the earlier seasons is the points where DGIndex Film % changes. usually one just after the eyecatch and then another about half way through the second half of the episode.

I am trying to figure out if DGIndex saves a log when you press "F5" I would like to get more detailed information. Like exactly which Frame # that percent changes. I assume this FILM % changes every time DGIndex sees a spot where it would need to insert or drop frames in order to "Force Film".

My idea was to look at the precise frame # where the film % changes and exactly how many frame #s were dropped or inserted, and then my hope is that the audio offset can be calculated from that.

this is how I have been fixing the earlier seasons manually:

A = NicAC3Source("F:\dbz\eng\001.ac3")
V = MPEG2Source("F:\dbz\eng\001.d2v")
f1=AudioDub(V, A).delayAudio(-0.242).trim(0,13166)
f2=AudioDub(V, A).delayAudio(-0.238).trim(13467,21692)
f3=AudioDub(V, A).delayAudio(-0.190).trim(21693,30000)
AlignedSplice(f1,f2,f3)

f1 is before eyecatch 100% FILM
f2 is after eyecatch 99.99%FILM
f3 is about half way through the remaining episode after the eyecatch 99.97% FILM

[Guest]

Film percent is averaged over time from the start of the project and so it cannot give you reliable frame-accurate information.

Film percent can change without inserting or dropping frames, as when pure 3:2 pulldown changes from soft to hard and back.

You have fallen down a rabbit hole with your sync obsession, in my view. Why don't you trust the authors? Is there some solid evidence that the coded delays are wrong by some objective criterion? What is your criterion for correctness?

[xekon]

What is your criterion for correctness?
My ears/eyes, there is a noticeable sync issue that is resolved when I adjust the delay. I have found several sound effects that happen 2-5 frames later than they are supposed to.


I could tell there was a noticeable audio sync issue ever since I started this project, but it was difficult to tell exactly what was going on.

So at first I thought it had to do with my decryption/demuxing process, so I started this thread.

After talking with you, I know now the correct way to handle the streams, and if I want to correct the sync issue it will be up to me to do it manually because the issue exists on the original dvds.

So I have been going about it manually. I will simply play the episode, and when it gets noticeable that it is out of sync then I do my best to figure out where that started and correct it.

So today as I was working on episodes. I would save a .d2v file for the next episode, and usually when its saving I watch the FILM% as its saving.

well as I fixed the sync on an episode today I noticed that the spots that it was out of sync matched the spots where DGIndex reported a change in FILM%.

AND just before writing this reply I noticed something else. Those spots where the FILM % changes are where it transitions from one CELL to the next.

However, when it transitions from Cell 1 to Cell 2 the film percent stays 100% (which makes since because I did not have to adjust the sync for the full first half of the episode until after the eyecatch), then from cell 2 to cell 3 the Film % changes, then again from cell 3 to cell 4.

Now I manually lined up this episode by ear before even noticing the coincidence of the FILM% changes being at those exact spots, then when investigating a couple minutes ago I discovered it was at cell transitions.

EDIT: The episode I sent you earlier was from the last season of the series, they did not have audio sync issues the same way, this seems to be something else entirely. on the one I sent you I applied a delay to the entire episode to correct it. On the earlier season episodes the amount of delay required changes drastically between cell 3 and 4

[xekon]

Ok so I took this one step further, using what you showed me earlier in this thread.

I used the ">" button, but I went all the way to 75% through the episode to the start of cell 4 and set the stop "["

When I do this, and use the delay reported in the outputed .ac3 file then that cell lines up the same way it does for the one where I did the entire episode as a single .m2v: "f3=AudioDub(V, A).delayAudio(-0.190).trim(21693,30000)"

So I am no expert and I am just guessing, but could the audio have ended before the video on cell 3? and when the audio for cell 4 got demuxed it was out of alignment with the video? because once I get to cell 4 I require a more positive delay value.

EDIT: If this is the case I will just use DGIndex per cell like this. so that each cell lines up the way it is supposed to, and then alignsplice() all 4 cells together.

when I make my .d2v files I will need to go beyond the cell by a couple extra ">" press or after using the reported delay I would have some silence at the end of each cell.

[xekon]

Well this is great. I will have a lot less work to do now :)

Is there already any guides on using DGIndex per cell and then Trim() and Alignsplice() them together like I will be doing.

If not I would be happy to write one. Im sure most DVDs dont have problems quite like these ones do. but some do I'm sure.

Edit: yep, was an open GOP, skipped a step in my decryption process that caused the issue, woops. So now I am back to just correcting the sync for the overall episode :) Ive got the first couple seasons audiodelay corrections done and I have encoded 42/291 episodes :)

[Guest]

So I am no expert and I am just guessing, but could the audio have ended before the video on cell 3? and when the audio for cell 4 got demuxed it was out of alignment with the video? It's possible yes, but it's also possible that you have an open GOP there. I always ask for stream samples.