Good.

One thing you can do is restore every track with the SC logo included.

Once you mux, just put on a delay of -21s or whatever the time is in the mkv. This will hide the SC logo from playing, and leave all audio and video untouched.

(2020-04-19, 05:05 AM)Stamper Wrote: Good.

One thing you can do is restore every track with the SC logo included.

Once you mux, just put on a delay of -21s or whatever the time is in the mkv. This will hide the SC logo from playing, and leave all audio and video untouched.

Yeah, that's one option. However, I'm thinking I may as well keep the SC logo at this stage. For the main stereo audio I'll quite likely be using LPCM anyway, which is headerless, meaning I could just concatinate the SC logo audio from the French BD to the modified main audio synced from any of the sources. Similarly, I believe AC3 is also headerless, so the same should presumably apply for the commentary and DVD stereo track if I make an AC3 encode of the SC logo audio to concat to the start of those. The French BD also has a 5.1 DTS-HD MA track, but everything after the logo is the wrong pitch and I'm not sure if the concat would work on that; I believe the DTS-HD MA header is already stripped when authoring a disc but I don't know if it's re-added by eac3to. I'd just change the pitch of the French 5.1 track but I'd rather avoid unnecessary manipulation of the main audio wherever possible, and the 24-bit DTS-HD MA on the US and UK discs is decent enough as it is (certainly better than the pitch-shifted French track), it just needs the older SC logo lopped off the start and the new one concatinated if possible.

I was originally going to just mux an MKV and then split it after the last frame of the 2018 logo (which runs from the start of frame 1 to the end of frame 516) but it turns out there isn't another keyframe until 10 frames later so that doesn't really work (it cuts 10 frames off the actual film, which breaks sync). I'm mulling over the best way to avoid this but I can probably just convert the frame numbers to timecodes (e.g. 517 / (24000/1001) = ‭21.563208333... sec a.k.a. 00:00:21.563208333) and use those instead to lop off the beginning of the audio files only. I'd not be keeping the video track split anyway, this would just be purely for convenience of splitting on the exact frame mark, but it seems that won't work for this particular video file.

Yeah, so I really don't think adjusting pitch in Audacity is going to work here. It sounds absolutely atrocious not matter what I do. It probably doesn't help that it sounded extremely tinny to begin with, but it's even worse after the pitch correction.

Hopefully the NTSC DVD audio isn't pitch-shifted so I can trim that down instead.

The 5.1 is easy all you will have to do is remux the one from the 2008 Blu-ray.

For the 2.0, I'm checking and will tell in a couple of hours. If it's good on the NTSC disc, then you will preserv from it.

EDIT: both tracks seems to be at the right pitch on the NTSC disc.

For your info it has 3 versions, DTS, DD and 2.0 DD + the audio commentary.

I would suggest to ditch any track from other versions not at the right pitch, and mux those, in addition to the 5.1 DTSHD at the correct pitch from the early BD

(2020-04-19, 12:43 PM)Stamper Wrote: The 5.1 is easy all you will have to do is remux the one from the 2008 Blu-ray.

For the 2.0, I'm checking and will tell in a couple of hours. If it's good on the NTSC disc, then you will preserv from it.

EDIT: both tracks seems to be at the right pitch on the NTSC disc.

For your info it has 3 versions, DTS, DD and 2.0 DD + the audio commentary.

I would suggest to ditch any track from other versions not at the right pitch, and mux those, in addition to the 5.1 DTSHD at the correct pitch from the early BD

Yeah I think that's precisely what I'm doing after discovering how crap and full of artefacts that re-pitching in Audacity ended up. It sounds OK with music but as soon as someone talks, the combination of artefacting from the original 25 fps re-pitch and artefacting from my re-pitching it back down is too much for me.

I've already taken the 5.1 from the UK BD since that's the right pitch, it's just the 2.0 AC3 and 4.0 AC3 commentary I need off the NTSC disc. I tried stretching my PAL tracks, trimming the audio for the old logo off the start of them, then concatinating the 2018 StudioCanal logo music on the start as same-bitrate AC3, and it worked a charm - this syncs it with the untouched French AVC video track without needing to remove the logo (otherwise I'd have to either remove too many video frames or too few, because of the location of the keyframes I can trim at). I did the same thing with the DTS-HD 5.1 tracks: trimmed the old SC logo off the UK BD then contatinated the 2018 SC logo from French BD onto the start of it without any transcoding whatsoever. EDIT: Forgot to change default audio when testing concatinated 5.1 DTS-HD MA, turns out it's added additional delay between the logo and main audio so there must still be a header on these that's being concatinated on; I'll need to find another solution. Wonder if I can maybe even insert a keyframe somehow (AviSynth?) to increase the accuracy of the trim and just lop the SC logo off, or something like that. It occurs to me that I may have needed to use the .dtshd extension in eac3to instead of the default .dtsma as Master Audio (e.g. as output by the DTS-HD MA encoder suite) does have a header; I think I can scrap the header by either trimming off 21 ms or maybe just by getting eac3to to demux to .dtshd in the first place.

As you will be syncing tracks from the NTSC disc, will you sync all of them? They should all play with the same delay.

Keep the Dolby Digital as it makes your project universal, and also the NTSC DTS as it may be differently encoded from the DTS-HD (which I don't trust).

The main problem to solve is the inability to trim off the precise number of frames for the two respective logos.

To sync the audio with the old logo to video with the new logo without re-encoding anything, I have to trim off precisely the right number of video frames' worth of time from each (e.g. in the 2018 French BD, 517 frames = 21.563208333 seconds). However, you can only trim video at keyframes, which are determined by the encoder, because all the resulting split parts need to begin on a keyframe to be decoded correctly. In my Snowpiercer resyn this is a non-issue because I got lucky and there happens to be a keyframe at precisely the correct place, but with Near Dark, *none* of the sources have a keyframe in the right place. This is less problematic for the audio sources because I can trim them more precisely using ms timestamps instead, but they still need to sync with the video track which I can't trim 100% accurately.

I therefore need to either:

1. Not trim the video track at all, and instead losslessly join appropriately encoded versions of the trimmed out 2018 SC logo audio to the start of each accurately-trimmed audio track (so it has the right audio for the new logo and syncs properly with the film itself), which is what I've been doing thus far, OR

2. Trim the video less accurately by working out where the earlier keyframe is and leaving some extra black frames in from after the SC logo, and trim & resync all audio to match that (removing the SC logo but leaving in some black frames due to the less accurate video trim).


I think I prefer option 1, but that potentially rules out using the old DTS track as I don't have a 5.1 DTS version of the 2018 SC logo audio, only one trimmed from the French DTS-HD MA 5.1 (which is pitched incorrectly apart from the logo), two AC3 ones I encoded at the appropriate bitrates to concat to the AC3 commentary and 2.0 main audio, and PCM ones to concat to the pitch-corrected lossless 2.0 tracks that I'm likely abandoning due to artefacting. Although, in fairness, I could probably encode a 5.1 DTS version of the logo audio if I really wanted.

I appear to be able to concatinate PCM (which makes sense because it's raw data) and AC3 (which is apparently headerless) without any issues, but concatinating DTS-HD Master Audio in either .dtsma or .dtshd containers seems to have failed, perhaps because the second part (the film audio) has a header on it. I've read elsewhere that this can be removed by eac3to by remuxing it with a -21 ms "delay", although I didn't expect it to be necessary in this case because the source for all the DTS-HD audio was Blu-ray Discs, which I thought had the header removed already (I guess eac3to maybe adds one back on). I'm about to try that and see what happens, but I'm not that hopeful because this seems to be significantly more than 21 ms out of sync (though I guess it might add more delay than that for the header after remuxing, idk).

No one cares about the SC logo. In fact if the music don't match, the better LOL, just sync the tracks with different opening logos.

(2020-04-19, 05:11 PM)Stamper Wrote: No one cares about the SC logo. In fact if the music don't match, the better LOL, just sync the tracks with different opening logos.

Never underestimate how chronically picky I am...

Seeing the new logo with the old music makes my brain scream. But it looks like I can't concat two DTS-HD MA files no matter what (from what I can tell thus far) so my only other option is most likely just cutting off the start of the French AVC track so there's no logo at all, but leaving in a few black frames due to where the keyframe cutting point lies. I would then make sure the audio all syncs to those black frames correctly so the whole thing works (e.g. if I've trimmed off the audio to start exactly in sync with the main video, and it turns out that the last keyframe is 3 video frames before that, then I could just mux the audio with a 125 ms delay (since 3 frames / (24/1.001) = 0.125 seconds).

EDIT: ... However, turns out the keyframes are quite far apart, so it can only cut to either include 2 frames of SC logo followed by 12 frames of black before the actual film, OR cut off 2 frames of the DEG logo starting (they're basically black frames anyway but are from the scan because there's grain etc. on them unlike the 100% black frames before that). This is... annoying.

One other small difficulty with all this is that I'm working with AC3 rather than PCM for some of the audio, and AC3 can only be trimmed to whole frames, each lasting 32 ms I believe (similar to H.264 video only being trimmable at keyframes). Resultantly, I can get pinpoint accuracy to 1 ms with PCM or DTS/DTS-HD MA but the AC3 might be out by about 11 ms, although that's only about 1/4 of 1 video frame so should be imperceptible in practice. Alternatively, I could decode the AC3 to PCM to be more accurate but it's probably pointless using so much extra storage space for the sake of 11 ms of sync accuracy.

I think you're complicating it. Simply remux the tracks, no need to touch them.

All you have to do is find one point in the film, find the same point in the DVD, calculate the difference in ms between the two points, and remux.