You_Too said:
I also think that overall, T2 needs a slight saturation increase as well as balancing out the pink skin tones.
I agree. T2 definitely needs less pink skintones.
You_Too said:
AntcuFaalb said:
Doctor M: Using ColourLike with every=1 will generate one histogram for the entire clip it operates on. If you open the histogram in a text editor, you'll only see one set of histogram values. It does not balance the color on a frame-by-frame basis. It only samples and updates the histogram "every" number of times.
I took a peek at the source code recently (it's in C++) and verified that this is the case. It's stupid behavior, but that's how it works!
If you want it to work on a shot-by-shot (or frame-by-frame) basis, then you need to generate histograms for each shot, manually. An example:
myclip = AVISource(...)
clip_one = myclip.Trim(0, 23)
clip_two = myclip.Trim(24, 47)
clip_one.WriteHistogram("shot1.txt", every=1)
clip_two.WriteHistogram("shot2.txt", every=1)
clip_one.ColourLike("shot1.txt", "shot2.txt")
Is there any way for you to change the source code and somehow create a new version that would work frame by frame? Or does that code you posted actually work frame by frame? (Even if it's slow)
The code I posted does not work frame-by-frame. I wrote a function to do frame-by-frame ColourLike using ScriptClip, but it's at home.
I'll write another one up now (at work now; slow day) and PM it to you, but I won't be able to test it.
Can you test it? We can post the final result here.
timdiggerm said:
Puggo - Jar Jar's Yoda said:
I figured the 30th anniversary would be a 6-month orgy
Holy cow! I feel like I'll some practice to prepare!
No. All you need is lessons from this kid: http://www.youtube.com/watch?v=zq5-6PkVGCg
Hey, it's me. said:
Jeez
Yes?
Lastly, if you discover that you're color-matching a film that was only globally tinted and not recolored shot-by-shot (e.g., a Blu-ray with a simple global teal-and-orange tint), then You_Too's method definitely works best.
It's also worth mentioning that a 50% color-match tends to blend better than a 100% one.
By this, I mean: run recolored.Merge(original) after the shot-by-shot application of RGBMatch.
Also, ContraSharpen (not really a sharpener) can help if color-matching a shot results in blown-out whites.
My preferred color-matching method at this point in time is g-force's RGBMatch.
I'm still holding out for something better, but in my experience so far, RGBMatch gets you at least 80% of the way there if you do it shot-by-shot.
Using RGBMatch frame-by-frame results in too much flicker and color balance fluctuations. I managed to get ColourLike to work on a frame-by-frame basis using ScriptClip and it exhibits the same kind of flicker and fluctuations, unfortunately.
stretch009 said:
AntcuFaalb said:
dvdmike said:
stretch009 said:
dvdmike said:
drngr said:
2 & 3 HDNet captures are now on Usenet and the bay by which buccaneers congregate (3 posted twice by mistake -- later torrent preferred but it's the same .ts).
Love the first comment.
I thought the HDnet versions were no good, low bit bate and some DNR I did get them off that place and posted here and I thought we all decided we wanted the wowow or canal +
The HDnet versions that were no good with low bit rate etc; were the highly compressed versions from rutracker. drngr just uploaded the uncompressed source files he found to usenet and TPB. I have both the compressed and uncompressed of Part II HDnet and the uncompressed looks WAY better than the compressed.
Interesting, they are falling off the internet as we speak
What do you mean?
A different way of saying "I'm downloading them", I guess.
Oh! I thought you meant that a DMCA takedown notice was issued and you were watching the individual RARs disappear from Usenet, one-by-one. lol!
Doctor M: Using ColourLike with every=1 will generate one histogram for the entire clip it operates on. If you open the histogram in a text editor, you'll only see one set of histogram values. It does not balance the color on a frame-by-frame basis. It only samples and updates the histogram "every" number of times.
I took a peek at the source code recently (it's in C++) and verified that this is the case. It's stupid behavior, but that's how it works!
If you want it to work on a shot-by-shot (or frame-by-frame) basis, then you need to generate histograms for each shot, manually. An example:
myclip = AVISource(...)
clip_one = myclip.Trim(0, 23)
clip_two = myclip.Trim(24, 47)
clip_one.WriteHistogram("shot1.txt", every=1)
clip_two.WriteHistogram("shot2.txt", every=1)
clip_one.ColourLike("shot1.txt", "shot2.txt")
dvdmike said:
stretch009 said:
dvdmike said:
drngr said:
2 & 3 HDNet captures are now on Usenet and the bay by which buccaneers congregate (3 posted twice by mistake -- later torrent preferred but it's the same .ts).
Love the first comment.
I thought the HDnet versions were no good, low bit bate and some DNR I did get them off that place and posted here and I thought we all decided we wanted the wowow or canal +
The HDnet versions that were no good with low bit rate etc; were the highly compressed versions from rutracker. drngr just uploaded the uncompressed source files he found to usenet and TPB. I have both the compressed and uncompressed of Part II HDnet and the uncompressed looks WAY better than the compressed.
Interesting, they are falling off the internet as we speak
What do you mean?
penguinofgreatness said:
Best of all, it's a DNR-free GOUT! I'd be interested to see if this is a practical idea to do for the whole movie.
Unfortunately, it isn't at this point in time.
You need a near-perfect image registration algorithm and, even then, you need to check shot-by-shot to see if the algorithm messed up, as even the good ones tend to mess up aligning small details, such as stars in a starfield.
g-force: Based on your experience, do you suspect that a YUV version would match better than an RGB version?
g-force said:
Sorry to be a downer.
Don't be sorry. You weren't a downer.
Thanks for the double-dose of reality. Recobbled was a bad idea.
OK. Recobbled has been scrapped.
g-force: Have you seen the following? http://originaltrilogy.com/forum/topic.cfm/New-colormatching-script/post/634651/#TopicPost634651
g-force: I agree. I'm fairly certain that I'm insane... sometimes. Maybe.
I didn't realize that the DNR in the 97SE is worse than GOUT. If that's the case, then I'll scrap that entire project idea.
Anyway, I plan to start capturing when my LX-900s return from Duncan of http://www.laserdiscservice.com/. Any day now...
I need ya'll to keep me level-headed. ^_^
Also, I forgot to mention that if we encounter a situation where it's like:
Shot N: 97SE
Shot N+1: GOUT
Shot N+2: 97SE
Shot N+3: GOUT
...
Then we'll just go with GOUT for that entire portion.
doubleofive said:
You realize you're going to have to pull from the GOUT for EVERY SINGLE WIPE in ANH and ESB, right? And most if not all of the model shots? That might get tedious/insane. At that point, you might as well watch the GOUT.
Yes, I do realize that. Thankfully, it only amounts to ~250 shots for A New Hope.
I completed IP reel #1 not too long ago and the assembly wasn't that bad.
The tedious part will be running MDeGrain1 on the GOUT portions to match-up with the quality of the 97SE shots. Looking for MDeGrain1 artifacts is not fun.
NeverarGreat said:
So for Recobbled, I assume you would use the GOUT as a starting point, then recolor it to match some previous laserdisc release or theatrical presentation, right? Now how would you get that to be better image quality than the GOUT unless you did image registration? Because I was under the impression that the only publicly available elements prior to 2004 were of lesser quality than the GOUT.
If image quality was the goal, why not just use the Blu-rays and do what Harmy is doing with Despecialized? I guess I'm just confused as to the ultimate goal of this project.
The primary source for Recobbled would be the best of the 97SE preservations. Excluding colortiming and framing differences, ~88.7% of the 97SE is completely untouched.
The 97SE preservations are higher quality than GOUT (very little DNR and EE; less dirt and grain), but are not so strikingly different that the transition from a 97SE shot to a GOUT shot would be noticeable.
That's the problem with using the '04 DVD or the '11 Blu-ray as a source. They're so high quality that you have to go full-Harmy for GOUT not to stand out. (Plus, I'm not interested in doing custom mattes, roto-scoping, etc.)
Harmy's project is great and I try to help out as much as I can with suggestions and seeding, but it's not for me. I may be speaking OTcom-heresy here, but I have to say that I don't like his colortiming. Also, I'm a dirty-little-purist and I can't stand that some elements from the SE (e.g., recomposited models, new wipes by Pacific Titles, etc.) are left in. I understand why he chose to keep those elements in and I agree that his decision to do so makes sense in the context of his project, but it still irks me. (No offense intended, Harmy! You do excellent work!)
So, in short: Recobbled would be an attempt to get the best quality version of the OOT on DVD using 100% original shots, no exceptions!
SilverWook said:
That poster has a real Frink vibe to it. ;)
"Frink Vibe" sounds like an excellent energy drink.
Ronster said:
Antcufaalb can I ask a favour for this thread please....
Primarily I think Retro Visuals Is key but I would also like you to update the first post with anything else that could be retro.
Which the only other 2 things I can think of is Retro Sound and music and retro plot ideas...
I changed the title and added a note to the OP.
Fixed?
NeverarGreat said:
Wow. I don't think that me and my single 1995 VHS of Empire will be much help to you. Though if you sent me screenshots and such I would happily look over them, I'm guessing you want someone with the sources to check against.
I'm not looking for sources to check against, actually. I just took the opportunity to list all the stuff I've acquired, lol. ^_^
I'm most interested in doing the "Star Wars: Recobbled" project I listed above. Another pair of eyes and a second opinion would go a long way with that one.
Video Collector said:
lpd said:
Im very interested, collector how do they hold up to retardtedted's project, picture and sound wise?
LOL, they probably don't hold up at all. Retardedted did an amazing job with noise-reducing, deinterlacing and stabilizing his NTSC sources. I'm sure a raw PAL capture can't compete.
Can you provide us with some details?
I might be able to NTSCify them and then run them through my Panasonic DMR-ES10 for improved stabilization.
Here are the StarWars-related VHS tapes I own:
(Everything not marked with a * is sealed.)
1978 PS78 Star Wars Bootleg*
1982 A New Hope Video Rental Library*
1990/92 A New Hope P&S "Your children and your children's children"
1990/92 The Empire Strikes Back P&S "Your children and your children's children"
1990/92 Return of the Jedi P&S "Your children and your children's children"
1992 A New Hope Special Letterbox Collector's Edition
1992 The Empire Strikes Back Special Letterbox Collector's Edition
1992 Return of the Jedi Special Letterbox Collector's Edition
1992 From Star Wars to Jedi: The Making of a Saga Special Letterbox Collector's Edition
1995 A New Hope THX P&S "Faces"
1995 The Empire Strikes Back THX P&S "Faces"
1995 Return of the Jedi THX P&S "Faces"
1995 A New Hope THX Letterbox "Faces"
1995 The Empire Strikes Back THX Letterbox "Faces"
1995 Return of the Jedi THX Letterbox "Faces"
poita said:
Confirmed that the VHD is *not* time compressed, but does have burned in sub-titles.
Interesting! I wonder if it was struck from the same master used for the 1982 VHS and 1985 CAV LD...
There are some other non-preservation projects I'm interested in tackling as well.
GOUT Shot-change Indices
I'd like to give back to the community a bit with a list of the [begin, end] GOUT shot-change indices. Most of this can be automated with an Avisynth script I wrote, but it'd have to be verified by human eyes.
A GOUT Release
I'd like to shot-by-shot recolor GOUT to resemble the home video releases of the 1980s; in particular, the 1985 Standard Play P&S LD. The aforementioned release is probably the most accurate "official" home video release, but exhibits some 1980s home video release flaws in the colortiming. Regardless, it seems to be what most people remember best. I'd shoot for an anamorphic DVD release of this. We can cleanup many of GOUT's non-DVNR flaws along the way, too.
Star Wars: Recobbled
An alternative to the project detailed above. I'd like to assemble the best pre-2004 sources for each shot (including unchanged shots from the 97SE) and make a better-than-GOUT DVD release of the trilogy.