The 1997 OT Special Edition Trilogy Preservation Standards Thread (* unfinished *) — Page 5

Okay, I'd like to put out feelers here to see if anyone else here, audio minded, would be willing to help keep this audio project afloat?

I don't want to see this project go belly up.  CapableMetal did some really nice work getting the raw audio out for folks to fiddle with.  I was hoping Eldonante would be able to have some free time to help out but he's asked to have the torch passed.

Even if there is anyone here that can put together a coherent and functional tutorial on how to process the audio getting it ready for the final phase, it would be greatly appreciated. 

I know each reel of audio has a bit of sound from the previous reel on it.   Do we have any solid time code for the theatrical version of the SE's that would help us sync the audio as close to the original as possible?

Any ideas, thoughts, or suggestions....please feel free to post them here.

Cheers!!!!

Hello all , long time away with real life stuff but got drawn back to the site due to the disney news.

Great to see loads of interest again in the 97 SE, can I help out at all?? I still have all my project files from the 'superset' and I still intend to finish it one day but real life does have a habit of getting in the way.

My intention was always to merge the various DVB sources and combining them with whichever stereo track was best (probably the TB / GKAR) and also syncing up an 5.1 laserdisc rip.

I think ESB is practically done except for some colour correction on ~13 frames, but things got tricky on ROTJ when I tried to get something good out of the blurry TB and heavily filtered GKAR, I'm just looking at some of my avisynth scripts which were intended to automagically fix this but they need a bit more work.

Something I've not seen mentioned is the various cropping and stretching issues in the DVB sources, from memory FLUNK has the largest frame, TB and GKAR are cropped at the sides and REIVAX also has a few lines cropped from the bottom, but all 4 sources line up nearly pixel for pixel (one of them has a sub-pixel shift to the side but I forget which. Also they don't have the correct aspect ratio (even when taking into account the incorrect encoding).

Maybe, I'll post a more comprehensive review of what I know about these sources when I get my main machine running (currently boxed up due to house renovations)

If there is a standard frame count to work to then I might have a look at finishing up the video and let others worry about audio stuff.

But maybe this has already been done or superior audio sources have been found?? I think there's a lot of stuff I'm not following in this thread due to several years away.

captainsolo said:

DJ and U2 are tackling n upscale and touch up to the TB versions after the Blu v2 is done of the GOUT.

Actually we're planning to release each 97 SE version along with it's original counterpart.

I'm still here, but working on a decent and colour-accurate video source. I managed to get a reasonable frame-accurate capture of the PAL laserdisc of ANH and averaged 5 captures to make it look a bit nicer, but its a laserdisc so there's only so much that can be done! All the broadcast versions, whilst mostly having superior picture, albeit with a more visible smear in parts, are missing frames, although I'm also looking into picking the best parts to make a better source. Please note that these are personal projects, so I won't be treading on any toes of other projects that may be happening!

I'm not sure there is any specific technical way to line the DTS up perfectly with any video source as we don't have a negative with the timecode markers on, or any technical means to perform a 'true' sync with the discs. The lead in/out on each reel doesn't give a specific reference point either. It seems to be a case of lining them up manually, and putting a slight overlap mix to make sure its seamless. Perhaps not perfect, but if each reel is lined up as correctly as possible with the video source then it should be unnoticeable. I used the '04 mix as a loose reference (and by loose, I mean just that, as dialogue and sound effects don't always match up due to it being an entirely different mix). I synchronised both ANH and ESB this way and it seemed to work effectively. I'm currently trialing a 24fps source instead of the usual 23.976fps; from the brief test I've run so far, no sample rate change or time manipulation is needed for the DTS audio to match up.

This is probably why 24fps works best...

    Film uses 24fps, and more recently, 48fps and even higher. Timecode in theatrical films is different and very simple — it counts every frame from number 1 to the total number of frames in the movie.

I'm keeping my eyes open over @ the 35mm Forum in case a set of the SE Trilogy shows up because that would ultimately be the best source for this entire project.  PROPS to any and all projects being able to make use of these DTS scores....I think it is a testament to this place that such work can even be done.  OT.com RULES!!!  :)  :)

Here is this info if it helps......

ANH

Six reels
125 minutes

TESB

Six reels
127 minutes

ROTJ

Seven reels
136 minutes

Movies were Scope(2.39:1).

For DTS audio synching, I'd guess we could use the time of each reel to figure out fairly accurately which frame each one ends on.  Just a thought.

bump

Jetrell Fo said:

This is probably why 24fps works best...

    Film uses 24fps, and more recently, 48fps and even higher. Timecode in theatrical films is different and very simple — it counts every frame from number 1 to the total number of frames in the movie.

That's exactly the logic I've been using since discovering that Blu-Ray supports 24fps. I've been piecing together a basic colour adjusted 720p24 version of ESB to try adding the DTS mix to. I'm going to try to place it without mixing adjoining tracks together, a straight cut from reel to reel if possible, it may take some extreme close-up editing to match it sample to sample to get it completely accurate!

_,,,^..^,,,_ said:

I thought to pass the audio from the VCR to my Sony MiniDisc recorder (because it has a really good A/D converter - 20 bit delta-sigma ADC - that I think is really superior to the simple 16 bit audio card of my PC) then feed its digital out to my PC. Is there something like AviSynth median I could use for audio?

I'm happy I decided to drop DScaler and instead use VDub to capture some laserdiscs - at the end, I captured 1 hour with NO frames dropped - just disable the preview... my video experience is slowly growing!

I haven't used my old MiniDisc recorder in many years, can you record to uncompressed PCM or does it compress to ATRAC? If it compresses it you should be better off using the capture card, as you will lose frequencies from your lower fidelity source. Averaging for sound surely does exist but I don't know how you would go about doing it as I've never done it! You would have to line up each recording sample for sample (or as close as you possibly can) for it to work properly, just as you do with multiple video captures, except there are far more samples/second in audio than frames/second in video. Random noise can change the shape of a waveform, making this potentially very difficult from a noisy source, but certainly possible in theory. It would just be easier to run a subtle NR pass with a negligible difference in perceived quality. Perhaps Audacity will do the job in this respect. I'd recommend having a bit of a play with different methods and see what works best, but I'd recommend getting some high quality shielded cables, if you haven't got some already, to reduce line noise from source to capture devices.

Incidentally my capture card keeps reporting no dropped frames in VirtualVCR, but still keeps missing 1 frame on all captures, which means I have to compare, then recapture and splice for that section. I'm a relative newbie to video work too, so its been a good way of learning, although frustrating at times because it means getting a frame-accurate capture takes a lot longer. I did manage to get a nice capture of the PAL laserdisc for ANH after getting a median of 5 captures. I'll give VDub a try and see if it gives better results!

How can I tell if my LD captures are frame accurate?  I've got my Japan LD captures of the S.E..  

It could give us another source to work with.

Any ideas? 

Jetrell Fo said:

How can I tell if my LD captures are frame accurate?  I've got my Japan LD captures of the S.E..  

It could give us another source to work with.

Any ideas? 

The way I've been doing it with my captures is to capture more than once, which was for averaging out noise. Normally dropped frames are reported by VirtualVCR, but there always seems to be one frame the slips under the net. Your capture card may well be capturing perfectly, of course!

The way I did it was to write down a frame number for a scene change every 5000 frames or so, and made sure each capture started from the same point by selecting a unique frame (like a frame on a THX logo with a unique piece of white dust) and setting a trim point in AviSynth. Do this for a point at the start and end and check each capture length in frames, and give a quick check for the scene change points, starting from the end. If you have 2 captures with an unreported dropped frame each, then at some point your numbers won't match in VirtualDub. Its then a case of finding the frame(s) that have gone and recapture that part of the film and splice it in with a simple AviSynth script and you should have a full capture ;)

EDIT: Just thought, you could trim from the same frames at the start and end of each capture, crop each one half way across the frame and stack them horizontally to check the scene transitions line up properly. Thats probably a faster method of doing it rather than checking each one individually.

^^^^

That's gonna take me some time.  My capture skills are probably pretty basic and my knowledge and usage of AVIsynth is far less basic...LOL

Jetrell Fo said:

^^^^

That's gonna take me some time.  My capture skills are probably pretty basic and my knowledge and usage of AVIsynth is far less basic...LOL

Hehe, I'll point you in the right direction and explain the script side of it then:

ClipA=AVISource("c:\PathTo\ClipA.avi").Crop(0,0,-360,0).Trim(245,65654)

ClipB=AVISource("c:\PathTo\ClipB.avi").Crop(360,0,0,0).Trim(327,65736)

StackHorizontal(ClipA,ClipB)

"ClipA="/"ClipB=" is defining what Clip A/Clip B is, just be sure to add any extra filtering you're doing to that clip on the same line (separated with . ), so it you wanted to separate the fields, for example, you just add '.SeparateFields() to the end of the line. In this case it simply loads the AVI file (assuming that's the format you've captured in) and crops one clip by 360 pixels right and the other 360 pixels left (just take the horizontal resolution of your captures and divide by 2, should work fine provided the heights are the same). The clips are Trimmed to start and end at the same frames as picked out in VirtualDub using your eyes ;)

StackHorizontal does just that, and because both clips are cropped 50% each way they should make a full image. Now just check it through in virtualdub and it should match perfectly, if it doesn't then you're missing frames.

Note: this is the most basic script I can think of, you'll likely end up running some more filters. SeparateFields() is an essential one to remember (as is Weave() to put them back together again!), especially when creating a manual IVTC script of all those NTSC laserdiscs ;)

If you have any trouble just PM me your scripts and I'll check them for you.

Are there any accurate sources for colour for the SE's? I know that the colours were tweaked for the 2004 masters (including the ridiculous level adjustments that washed out the saber cores/laser blasts/random shots), and shots given colour tints.

The way I understand it (and vaguely remember it in the cinema) is that the SE's have a colour-timing closer to the THX OT. I've been colour correcting TESB from the Blu-Ray and I've come up with some pretty good results so far, but with only analogue and broadcast versions to go by, which need colour corrections themselves, its been difficult to discern what is accurate and what isn't from memory.

I've also identified shots where white levels have been dialed down and was shocked to see how awful it is in places. The final shot of ANH is clearly a victim of this, along with the lightsaber cores/laser blasts/explosions/various random shots in every film.

Is there anything that can be used a reference?

I saw Jetrell Fo posted about film being 24fps, though I still wanted to post this.

Me and DJ are going to use this audio in our 97 SE project, and DJ had already synched the first reel of ANH but told me it was very hard. Then I got that idea that it might need a conversion from 24 to 23,976 fps, and I tried it myself and it was an almost perfect match with the digital broadcast version we're working with. (TB version, converted from 25 to 23,976 fps)

Though the software I used to convert the audio from 24 to 23,976 (BeSweet) is old and felt very outdated, since it forced every mono file to be converted to stereo during the conversion.

I also wonder what would be the easiest way to convert this from 44100hz to 48000hz? Or does a blu-ray accept audio being 44100 if we'd make a DTS track from this?

EDIT: Found out that eac3to could do this conversion properly. It was hard to find out how to make a script that would convert all audio files at once though.

You_Too said:

I saw Jetrell Fo posted about film being 24fps, though I still wanted to post this.

Me and DJ are going to use this audio in our 97 SE project, and DJ had already synched the first reel of ANH but told me it was very hard. Then I got that idea that it might need a conversion from 24 to 23,976 fps, and I tried it myself and it was an almost perfect match with the digital broadcast version we're working with. (TB version, converted from 25 to 23,976 fps)

Though the software I used to convert the audio from 24 to 23,976 (BeSweet) is old and felt very outdated, since it forced every mono file to be converted to stereo during the conversion.

I also wonder what would be the easiest way to convert this from 44100hz to 48000hz? Or does a blu-ray accept audio being 44100 if we'd make a DTS track from this?

EDIT: Found out that eac3to could do this conversion properly. It was hard to find out how to make a script that would convert all audio files at once though.

eac3to is an excellent tool ;)

I've done near-perfect synchs of all three films to 24fps. It was the easiest way, as they do match the film frame rate. I've also managed to synch them by adjusting the playback sample rate to match (roughly, but perfectly adequately) to 23.976 in my first attempts.

Blu-Ray format accepts nothing less than 48KHz, but interestingly enough does support a true 24fps format in 720p and 1080p, in addition to the usual 23.976fps.

Have you reinserted the missing frames into the TB broadcast you're working on?

Synch should be very easy if you have, I managed to synch all three films in less than 6 hours to a frame-accurate video at its native 24fps frame rate, then just ran a sample rate conversion to 48KHz when they were all done. If you haven't then you may have to adjust/mix the DTS tracks in places to match it; I know that in the TB broadcast of Empire most of the missing frames are in the middle of reels, something like 12 frames out of the 15 or so that are missing. The rest of the time they are missing frames around reel changes, which should be much easier for you to work around should you choose that path.

CapableMetal said:

Have you reinserted the missing frames into the TB broadcast you're working on?

Synch should be very easy if you have, I managed to synch all three films in less than 6 hours to a frame-accurate video at its native 24fps frame rate, then just ran a sample rate conversion to 48KHz when they were all done. If you haven't then you may have to adjust/mix the DTS tracks in places to match it; I know that in the TB broadcast of Empire most of the missing frames are in the middle of reels, something like 12 frames out of the 15 or so that are missing. The rest of the time they are missing frames around reel changes, which should be much easier for you to work around should you choose that path.

Actually DJ is the one who is synching them, but we don't even know where all those frame drops are? Obviously some are at the reel changes and since your audio is divided into reels it should be easy to synch them reel by reel, but I didn't know they were missing frames in the middle of reels. It would be very nice if you'd point out where.