Streets of Fire (1984) - Best audio?

[pipefan413]

Right. So. I'm still up to my tits in Terminator 2 and am fiddling with that Anatomy of a Dewback featurette that nobody on the planet bar me actually cares about at this stage (especially since I've now decided that the version on the 2011 BD is basically correct apart from the annoying fact of it being 4:3 letterboxed to a kind of dodgy not-quite-but-nearly-1.66:1) BUT NONETHELESS



I believe what I now have basically amounts to the following:




VIDEO:

1. 2013 UK Second Sight Blu-ray Disc: AVC encode from old master (very green, not very great fidelity)
   
   
2. 2017 US Shout! Factory Blu-ray Disc: AVC encode from new master (colours signiificantly different but still possibly questionable, encoded by Shout so probably iffy in places, but looks like a significant fidelity upgrade from anything beforehand)
   

I do not have the German Mediabook from 2019 because it is bloody expensive and I honestly don't think it's better overall than the Shout! Factory release, because it appears to be way over-cropped to my eye, even if the encode could theoretically be handled better than the S!F one. I dunno.




AUDIO:

1. 1996 US Universal LaserDisc: 16-bit LPCM 2.0 (now... weirdly, this is shown on lddb.com as straight stereo, but I'm convinced this is wrong because this is a Dolby Stereo film which means it should be inherently and irreversibly matrix encoded from 4 channels to 2, decoding back ideally to L|C|R|S but in practice to whatever your modern decoder is capable of, which is probably 5.1 / 7.1 or something)
   
   
2. 2013 UK Second Sight Blu-ray Disc: 16-bit LPCM 2.0 (Dolby Stereo matrix encoded L|C|R|S)
   + 16-bit DTS-HD Master Audio 5.1 (I guess upmixed from the 4.0 as discrete channels instead of being 2.0 encoded, with a low pass filter to derive LFE etc.)
   
   
3. 2017 US Shout! Factory Blu-ray Disc: 24-bit LPCM 2.0 (Dolby Stereo matrixed L|C|R|S) which supposedly suffers some dropouts, which I've yet to verify
   + 24-bit DTS-HD Master Audio 5.1
   
   
4. 2018 US Shout! Factory Blu-ray Disc Steelbook (in the post): 24-bit DTS-HD Master Audio 4.1 mix from 70 mm Six Track Dolby Stereo master (L|C|R|S|LFE)
   

It'd be cool to have the German disc to more or less complete the set but I don't think it's worth it at that kinda cost when the cropping looks wrong to me anyway. What would be valuable is the video from the LaserDisc, partially to see how it looks compared to the older BDs like the Second Sight one, but mostly to use to determine the locations of missing frames to help map out a more accurate sync for mapping the LD audio to BD video.

I'll probably start throwing the Second Sight and Shout! Factory video into AviSynth this weekend; I already made crappy 64 kbps encodes to do just that without having to stress the hell out of my video card by scrubbing through full bitrate AVC frame by frame. I'll still verify that the transcode didn't drop any frames afterwards just in case by supplanting the transcode in AviSynth with the original AVC and making sure it lines up after all the edits are made to sync the two video streams.

[pipefan413]

OK so I lined up the 1996 LaserDisc to the 2017 Blu-ray Disc by removing ‭1,385,201‬ samples (31.410 seconds) and it actually stays reasonably close for the most part but gets ahead by about 0.113 seconds by the end of the side (about half way into the film). That isn't that surprising as I'd expect the older master used for the LaserDisc to be missing any number of frames and there'll have been 2 or 3 reel changes by that point, but I can't really check *exactly* where they are without the video off the LaserDisc. I guess what I'm going to do is to just go through looking for quiet moments (sadly few and far between in this film!) to loop/extent slightly such that sync drifts ever so slightly to let the more frame-complete Blu-ray video stream catch up with the LD audio.

HOWEVER... with this film, I think it is reasonably likely that the LaserDisc and the older Second Sight Blu-ray are from the same master, and if so, will probably have the same missing frames. If that turns out to be the case, I shouldn't even need the LaserDisc video to do this accurately, because what I can do is...

1. Sync the 1996 Universal LD audio to the 2013 Second Sight BD video (which should line up better than the Shout! Factory video, since that one's from a different scan)

2. Sync the 2013 Second Sight video with the 2017-2018 Shout! Factory video in AviSynth to locate missing frames

3. Use locations of missing frames to target extensions to LD audio to sync to Shout! Factory video

Doing it this way also means I can do syncs for both versions of the video, as well. That's useful!

So I started doing exactly that, and found that there were an extra 24 frames of black on the start of the Shout! encode. So I added 24 frames to the start of the Second Sight video. And then I skipped through them side by side... still in sync after reel 1... still in sync after reel 2... and so on... weird, Terminator 2 was all over the place... holy crap I'm at the end of the film! The Blu-rays are *exactly in sync* apart from a 1-second (or more accurately, 24 frame) offset at the start! The only difference apart from that is that the Second Sight one ends with a PG rating card but the Shout! Factory one ends with a Universal Studios logo and that "Visit Universal Studios" card I've seen a couple of times elsewhere (can't remember which films).


In the meantime, here are some spectral analyses for the various audio tracks...


70 mm Six Track Dolby Stereo mix from 2018 Shout! Factory Blu-ray Disc (Steelbook):




35 mm Dolby Stereo mix from 2017 Shout! Factory Blu-ray Disc:




35 mm Dolby Stereo mix from 2013 Second Sight Blu-ray Disc:




35 mm Dolby Stereo mix from 1996 Universal LaserDisc:




Incidentally, the LaserDisc audio was originally longer than the others because it plays the audio from two trailers! At least one of them seems to be this one:

https://www.youtube.com/watch?v=lEOvn2IaLMM

[HippieDalek]

The only difference apart from that is that the Second Sight one ends with a PG rating card but the Shout! Factory one ends with a Universal Studios logo and that "Visit Universal Studios" card I've seen a couple of times elsewhere (can't remember which films).

The laserdisc also has the "Visit Universal Studios" card before it plays the two trailers.

Glad you managed to get them synced better. I tried my best but for some reason Audacity was sometimes trimming the silence at the start of the tracks. It's such a shame that Davinci Resolve doesn't output 44.1 kHz as it's such a good way to edit audio otherwise.

[pipefan413]

The laserdisc also has the "Visit Universal Studios" card before it plays the two trailers.

Oh, interesting! Did it have the PG rating card too, or only the Universal thing? The Shout! has no rating card, whereas conversely, the Second Sight only has the rating card but no Universal Studios bit.

Glad you managed to get them synced better. I tried my best but for some reason Audacity was sometimes trimming the silence at the start of the tracks. It's such a shame that Davinci Resolve doesn't output 44.1 kHz as it's such a good way to edit audio otherwise.

Oh, no, to be clear: I'm by no means trying to drop a big jobby on your work, I haven't even tested it out yet! My logic is I should try to do it entirely on my own first, then I'll be curious to see if our results are similar. Just means any discrepancies in what you've done won't colour my own decisions or whatever, right? And I haven't actually got it 100% yet, because... well, I'm about to go into detail, so buckle up if you're really bored. Here I go...


The calculations are as follows to sync the Second Sight BD to the Shout! Factory BD:

Code:

2013: 00 + 134931 + 000 frames
2017: 24 + 134931 + 349 frames = 135304 frames in total

24 frames at 48 kHz =  24 / (24000/1001) * 48000 =  48048 samples
349 frames at 48 kHz = 349 / (24000/1001) * 48000 = 698698 samples

Theoretically, Second Sight audio should last
134932 / (24000/1001) * 48000 = 270,131,862‬ samples
but it's actually 270,296,160 samples long for some reason

270,296,160 - 270,131,862‬ = 164,298‬ samples extra
so I'm only expanding the end by
698,698 - 164,298 = 534,400 samples

in other words, the Second Sight audio needs
48,048 samples added at the start, and
534,400 samples added at the end,
to be precisely the right length for the Shout! Factory video stream.

Having done that, it's exactly 270,878,608 samples long, which is precisely the duration of 135304 video frames (the length of the Shout! Factory video).

Observations re. Second Sight audio:

1. If you zoom in and look at it up close, it's *extremely* similar to the LaserDisc waveform. As in, visually, the waveforms (apart from small sync discrepancies) look almost identical. This is absolutely not true of the Shout! Factory audio, which looks significantly different.

2. On the other hand, though, it's also way more clipped / compressed than the 2017 Shout! audio; the waveform is clearly hitting the ceiling throughout, if you look at them side by side. This is unfortunate. Only the Second Sight disc has this problem; the 1996 LaserDisc and 2017 Shout! BD are both fine in this regard.

Conclusion: I think Second Sight used the exact same audio recording as the 1996 LaserDisc and just adjusted the sync a bit, maybe to account for additional frames recovered between 1996 and 2013, then compressed the crap out of it. By contrast, it appears that Shout! Factory have re-recorded the audio (since the waveforms are very different up close).


In addition to this, correcting by the exact number of audio samples to offset the difference in frames doesn't align the sync as well as you might think. Sync is noticeably different from the Shout! audio for whatever reason; in general, the Shout! audio falls a little over half a frame later than the Second Sight audio (a bit over 21 milliseconds) and it's pretty consistent from start to finish. That said, we're talking about not much more than a 50th of a second here, so it isn't going to be noticeable, regardless of which is actually more precise. Honestly, it's anybody's guess. Shout! screws up their video a lot and supposedly their audio on this film has dropouts in it, but I've not listened to it all the way through so I wouldn't know. What I do know is that Second Sight's audio is really compressed compared to either the Shout! disc or the old '96 LaserDisc, but that doesn't necessarily tell me anything about sync.


Anyway, what this means is that so far I have the following audio tracks in sync (give or take half a video frame's worth) with the Shout! Factory video track:

70 mm Six Track Dolby Stereo 4.1 from 2018 Shout! Factory BD (Limited Edition Steelbook)
35 mm Dolby Stereo 2.0m from 2017 Shout! Factory BD (Collector's Edition)
35 mm Dolby Stereo 2.0m from 2013 Second Sight BD

(I'm ignoring the 5.1 tracks for now because they seem pretty pointless)

And it theoretically shouldn't be *that* difficult for me to now sync the final track, the 35 mm Dolby Stereo 2.0m from the 1996 Universal LaserDisc, possibly with patching where required from the Second Sight track since it appears to be from the same recording. Frankly, it's probably close enough after what I've done with it already, but it does start to drift a bit around the first reel change (even then, it's only about 40-odd ms out, and I think after that it stays fairly solid) so I should be able to get it a wee bit more accurate, assuming the Shout! audio is actually accurate in the first place. I'll be keeping it 44.1 kHz until the very end of the process, then I'll resample it up to 48 kHz for Blu-ray Disc compatibility and possibly encode to DTS-HD Master Audio with a matrixed stereo flag so that you can either use the linear PCM at 44.1 kHz for the least messed with audio compared to the LaserDisc source, or a more modern version with automatic signalling to your AVR to decode the audio as surround and all that jazz.

[HippieDalek]

Oh, interesting! Did it have the PG rating card too, or only the Universal thing? The Shout! has no rating card, whereas conversely, the Second Sight only has the rating card but no Universal Studios bit.

I can't remember off the top of my head. I'll check for you once I'm done ripping LDs and I've got my player back downstairs and hooked up to the TV again.

Oh, no, to be clear: I'm by no means trying to drop a big jobby on your work

Ha ha, don't worry, I didn't take it that way

[BDgeek]

WOW, Very detailed info, awesome! Thanks a lot pipefan413!

Thanks again to HippieDalek for the great help securing the LD and ripping it.

And thanks to The Aluminun Falcon for the Shout's 4.1 70mm track!

After pipefan is done with the LD track, with the plethora of audio tracks available, I think we can definately cross this one out of the list for preservation : )

[pipefan413]

Getting this right is proving quite difficult because this film's audio never lets up! There's specifically a huge amount of music throughout almost every moment, so I don't want to be making adjustments to sync in the middle of a song, since the rhythm will make the adjustment too obvious and I don't want it to be noticeable. There's also pretty much no silence in the entire track, which makes it harder because I can't just add a bit of silence or near-silence to add time where needed. So I'm probably looking at patching from other tracks like the Second Sight one instead, which is something I'm new to doing so will have to figure out a whole new workflow for. I've also still got the Terminator 2 stuff to do at some point, so I'm probably going to be flitting back and forth between the two. And I should really be translating Hachiko but I'm doing this stuff instead...

Again, I could probably just leave it as it is but I'd prefer to get the accuracy up a bit! I don't think it's massively off and a quick cursory check of the waveform on @HippieDalek's sync has me thinking it's not that different to what I've done thus far myself but neither that or my WIP are dead-on yet. It's pretty much exactly in sync in some places but with those adjustments (both @HippieDalek's and mine) there's still roughly about 30-40 ms or so of a discrepancy depending exactly which moment of the track you look at, the sync of the actual released tracks drifts and isn't matched between them as I found when working on Near Dark so it's never going to be "100% accurate" anyway). I mean, realistically, we're talking about maybe like a 20th of a second of inaccuracy here so it's probably fine, but I'd still rather get it as close as reasonably possible. It looks like the 70 mm 4.1 track and the Dolby Stereo 2.0 sync on the Shout! releases are actually pretty close to one another sync-wise, at least, but they're not identical.

[BDgeek]

Thanks for the heads up. Looking forward to it!

BTW, I'll get you the T2 R4 AC-3 track.

[pipefan413]

Sorry, you'll have realised by now that I can't focus on one thing at a time so it takes me ages to finish anything. This is particularly true if I run into something messy and don't want to half-arse it. For instance, trying to resync this film without video reference.

In a fairly exciting development (for me at least) I have now bought my own LD player along with a handful of films for it, which were actually the main reason I finally pulled the trigger. But one of the other things I'd like to do is cap the video for STREETS OF FIRE to use as reference for this resync.

Now, I have found a copy of a different SOF LD, but not this one. And the one I've found is fairly unreasonably expensive because I have to either ship from Japan or buy it in a set of 8 with 7 others I don't have any need for. I would be less hesitant if I hadn't already spent more than I can realistically afford on hardware lately. So really the options for getting the video off this are probably as follows:

1. Get @HippieDalek sorted for really crappy basic video capture and hope that he caps the LD he's already got
2. Find and buy the exact same LD for my own collection and cap that
3. Get @HippieDalek to send me this one (I'd obviously pay postage for the round trip, preferably tracked to avoid anything untoward) and cap it on my end, but it'll likely get a bit bashed with all the shipping back and forth so maybe not a good option
4. Just buy the alternative LD and do a whole new cap of that with both video and audio, and resync that instead... or perhaps do that first and then try to align the original cap to that resync. That'd be nice because it would give us yet another option, but it'd be kinda expensive for me! XD

I'm already just about bursting with excitement at the prospect of capturing THE EXORCIST JP LD video to assist resync of the analogue mono (I'd be capping the audio again myself at the same time) as well as what may be a first time cap of NEAR DARK's Ultra Stereo mix off its US LD, again along with video to help resync. I just hope my chain turns out to be capable of bit perfect capture!

[Doctor M]

Lots of tracks, but do we have a "winner" in the pile? I'm still considering the Steelbook based on reviews.
Looking at the spectral analysis, it seems the 2013 and 2017 tracks are victims of the loudness war with the Steelbook audio being closer in dynamic range to the LD.