Proposal: The James Cameron Project

[The Aluminum Falcon]

Fairs points. See this is sort of the problem I am having and probably will have with this whole project. Do I skew close to print examples in the regrade or do match the color as best as possible but tweak to be more like a modern home video transfer? So do I push for colorful, contrast, etc or flatten the picture? I'm hoping to get sort of a "half-way point", whatever that might be. I did a tiny bit of tweaking but nothing major, I'll post some more hopefully this weekend.

First of all, sorry to hear about the emergency! Real life stuff always comes first, thanks for taking the time to respond. But hope all is well.

re: T2, my personal opinion for these sorts of regrades would be that you should always go for a print-type contrast (maybe with the highlights qualified/retained and a little more shadow detail, to be fair). I totally respect that some people like the lower contrast look they're used to, but, frankly, until home video came into being in the 80s, there wouldn't be really any reasonable expectation of seeing something with the low contrast of an IP unless you flashed the film, which would be a very specific creative choice.

There also exists a danger that an in-between compromise between this and film contrast just isn't an aesthetically pleasing look. YMMV.

To go to a broader topic, commercial modern restorations are also starting to avoid the low contrast IP look and get back to a more authentic filmic contrast. The latest restorations of Double Indemnity; The Good, the Bad, and the Ugly; Casablanca; The Apartment; Thief; Pulp Fiction; and the Indiana Jones movies all have higher levels of contrast then previous restorations, presumably to better emulate the generational contrast that a DP would reasonably expect and for which they would consequently set up their lighting.

Why do all this? Possibly to honor the filmmaker's intentions. But probably... because it looks better and helps the eye focus on what's important. When you increase the contrast on your favorite movies shot on 35mm, you normally find that what should be lit stays lit and meaningless detail falls to the wayside because, again, this would be what would naturally happen when a print was struck; phrased another way, it was the DP's job to make sure crucial visual information was sufficiently exposed to survive the generational process (even without fancy digital masking or windowing ).

Sorry that was a bit of a ramble, again this is all 100% my opinion, pls feel free to differ. In your specific case @PDB , I've greatly enjoyed your releases like Die Hard 3 that emulated a print level of contrast.

That being said, if you want to proceed with this lower amount of contrast, then I'd say you should probably neutralize the extreme highlights and shadows by desaturating them. You wouldn't really see many official home video releases where an image is completely veiled blue, green, or yellow. As would happen naturally with a print, the highlights/shadows usually are neutral to give the eye a better sense of color.

[Only exception is some early digital color grading which veered towards blanket tints, but, to be honest, this isn't as popular a practice nowadays.]

EDIT: On the topic of James Cameron, word is that in 1986, he collaborated directly with the color labs in charge of striking the main prints of ALIENS because he wanted more control than usual over the grade. So, to that point, James Cameron would be a director I'd trust to fully understand the generational process associated with striking film. Why he didn't pay more attention to his home video transfers... who knows?

[zoidberg]

So the story goes Cameron would bring his own TV set to home video mastering sessions to see how the transfers looked on a regular set as opposed to the studio calibrated monitors. I do think the inherently different look on home formats is down to using interpositives, the colour could only be pushed so far.
He was always super involved in the lab processes and would be on their case to keep the printing on target and would regularly run tests to see which lab was the best at any given time for super 35 work. He also paid personally for the 70mm prints of Titanic and Truel Lies to be struck.
Good luck with this project PDB, I know these films have been a passion project of yours for a while. I think I still have a test render of one of your T2 regrades

[LucasGodzilla]

Sorry had a family emergency and wasn't able to respond right away. I wanted to get deeper into this conversation but you will all have to excuse the brevity, got some things to work on.

Sorry to hear about that. I hope your family situation improves.

I tried for months to get autooverlay to work on a few projects but it never seemed to come together for me. I got close for some things but never close enough to release anything it produces. That's not on the app but on me. Clearly it works but I can't get it to work.

A shame. I also noticed the UHD also had much smoother highlights that actually rolled off rather than harshly clip, though I suppose there's little point in dwelling on that if you can't get the auto-overlay to work.

Trying to understand this better LG.

Is what you are seeing color in the highlights or the highlights pulled to much? I compared your examples versus my originals and I can see the differences. It looks like, for a quick example, the glass in bar over Arnold's right shoulder goes from bluer to redder and over Hamilton's right shoulder the glass goes from bluish to whiter/yellow. Also looks like less saturation and flatter (ie less contrast)?

So are you pointing towards highlights being "whiter" or white balanced?

Basically the same qualms as what TAF had, which is that the saturated highlights give the regrade a bit of an odd veiled look.

What I tried in my test was simply to desaturate the image towards the light with the tonemapping (I think the yellow bit is just a trick of the eye unless Resolve does something I'm not aware of). As a result though, yes, the image does look flatter and less saturated since the brighter parts of the image are rolling off to neutral, hence my added statement that you'd have adjust the grade a bit to compensate.

As TAF already said, some added contrast would be beneficial to the image. In Resolve for this test, I essentially just bumped up the gain a bit and tweaked the tonemapping settings a smidge to get this.



Whether or not it's considered an improvement is up to you though. Given how some of the blues turn a different shade post adjustment, you'd probably have to push the colors harder depending on the scene at hand.

[The Aluminum Falcon]

So the story goes Cameron would bring his own TV set to home video mastering sessions to see how the transfers looked on a regular set as opposed to the studio calibrated monitors. I do think the inherently different look on home formats is down to using interpositives, the colour could only be pushed so far.

Oh wait if Cameron used a home CRT, then that could go a long way to explaining the magenta look of the T2 transfer. PDB, have you tried putting the unaltered transfer on the CRT you use for your CRT projects?

[zoidberg]

The colours on the earlier T2 master (ie the Live or UE master) were much better than the EE master. Still not accurate to the theatrical look but not utterly borked like the EE

[alexp2000]

Sorry had a family emergency and wasn't able to respond right away. I wanted to get deeper into this conversation but you will all have to excuse the brevity, got some things to work on.

Thanks for the feedback guys. I will say those pics are from a rough first pass of auto color matching. I still have some global settings to add and then go through it shot by shot.

Interesting work PDB - really like the upscaling on the T2 images (that master certainly needs some upscaling!) The colours look a bit off though, looking forward to seeing how that one develops.

Thanks. In what way do you think the colors look off? Like just don't like or look wrong? I ask because if I can fix something I will.

Firstly hope everything works out ok PDB

I think others have covered the lack of contrast better than I could put it. For me, the colours just look a bit 'strong' in some shots - particularly the one of Arnold looking up with his face damaged and Sarah smoking. From what I've seen of the 35mm release that's out there you've got the right sort of colours but maybe a bit over saturated. I guess it might be down to the limitations of the Blu-ray?

[Stamper]

This is all ace, but I must join the general consensus re: T2, I also think some of these shots looks off. Especially the mini-gun one.
Also, the Sarah interrogation was never green in any 70mm or 35mm projection I saw.
The bar scene however was total blue, this was the one that was completely changed on home-video.

If you use the 2015 Blu-ray as base, I don't think lots of work would be needed, beyond matching to the 35mm prints.

Personally, I prefer the 35mm look, it's now my go to copy.

Re: T1 I can send you wav of all the files I used on my purist collection project.

Titanic : you might want to color match the Blu-ray to the original NTSC DVD (there's also an open matte of this one on the russian site).

[LucasGodzilla]

I tried for months to get autooverlay to work on a few projects but it never seemed to come together for me. I got close for some things but never close enough to release anything it produces. That's not on the app but on me. Clearly it works but I can't get it to work.

Bit of a follow-up actually, stumbled across a fairly old post Random.next made that managed to sync up these very masters it seems for at least some experimental screenshots.

I assume you've already looked at these back in the day and more recently for this project, but I thought it'd be interesting to quote the results here in case some reading this haven't / forgotten about its existence since the time it was posted...

New lionsgate transfer combined with 2017 remastered with AutoOverlay:
Hybrid vs BD vs Remastered
 

 

As well as a dump of the script he used to get those results...

Code:

NEWbase=AVISource("M:\MEDIA\T2\NEW.avs")
OLDbase=AVISource("M:\MEDIA\T2\LIONS.avs")

NEW=NEWbase.ColorYUV(off_y=0, gamma_y=0, cont_u=0, cont_v=0).ScriptClip("

tr=2
count=tr*2 + 1

lumaMedian=0.0
chromaMedianU=0.0
chromaMedianV=0.0
ownY = OLDbase.YPlaneMedian() - YPlaneMedian()
ownU = OLDbase.UPlaneMedian() - UPlaneMedian()
ownV =OLDbase.VPlaneMedian() - VPlaneMedian()
for ( i = -tr, tr, 1) {
   curY=OLDbase.YPlaneMedian(i) - YPlaneMedian(i)
   curU=OLDbase.UPlaneMedian(i) - UPlaneMedian(i)
   curV=OLDbase.VPlaneMedian(i) - VPlaneMedian(i)
   lumaMedian=lumaMedian + curY
   chromaMedianU=chromaMedianU + curU
   chromaMedianV=chromaMedianV + curV
}
lumaMedian = (lumaMedian/count)
chromaMedianU = (chromaMedianU/count)
chromaMedianV = (chromaMedianV/count)
ColorYUV(off_y=0, off_u=chromaMedianU, off_v=chromaMedianV)
")

OLD=OLDbase.ColorYUV(off_y=0, gamma_y=0, cont_u=0, cont_v=0).ScriptClip("

tr=2
count=tr*2 + 1

lumaMedian=0.0
chromaMedianU=0.0
chromaMedianV=0.0
ownY = NEWbase.YPlaneMedian() - YPlaneMedian()
ownU = NEWbase.UPlaneMedian() - UPlaneMedian()
ownV =NEWbase.VPlaneMedian() - VPlaneMedian()
for ( i = -tr, tr, 1) {
   curY=NEWbase.YPlaneMedian(i) - YPlaneMedian(i)
   curU=NEWbase.UPlaneMedian(i) - UPlaneMedian(i)
   curV=NEWbase.VPlaneMedian(i) - VPlaneMedian(i)
   lumaMedian=lumaMedian + curY
   chromaMedianU=chromaMedianU + curU
   chromaMedianV=chromaMedianV + curV
}
lumaMedian = (lumaMedian/count)
chromaMedianU = (chromaMedianU/count)
chromaMedianV = (chromaMedianV/count)
ColorYUV(off_y=0, off_u=chromaMedianU, off_v=chromaMedianV)
")

config=OverlayConfig(subpixel=2, branches=1, aspectRatio1=2.3, aspectRatio2=2.45, minSourceArea=70, minOverlayArea=70, minSampleArea=1000, angle1=0, angle2=0.08, debug=false)

OverlayEngine(NEW.ConvertToY8(), OLD.ConvertToY8(), editor=false, upsize="BilinearResizeMT", downsize="BilinearResizeMT", maxDiff=25,  maxDiffIncrease=3, maxDeviation=0.5, backwardFrames=3, forwardFrames=6, configs = config)

debug=false
luma=OverlayRender(NEW.ConvertToYV24(), OLDbase, debug=debug, colorAdjust=2, matrix="",  opacity=0.5, gradient=50, lumaOnly=false, upsize="Spline64Resize", downsize="Spline64Resize", mode=4, width=1920, height=840)
chroma=OverlayRender(NEW.ConvertToYV24(), OLDbase, debug=debug, colorAdjust=0, matrix="",  opacity=0.5, gradient=100, lumaOnly=false, upsize="Spline36Resize", downsize="Spline36Resize", mode=4, width=1920, height=840)
asOld=luma.Overlay(chroma, mode="chroma")
luma=OverlayRender(NEWbase.ConvertToYV24(), OLD, debug=debug, colorAdjust=1, matrix="",  opacity=0.4, gradient=50, lumaOnly=false, upsize="Spline64Resize", downsize="Spline64Resize", mode=4, width=1920, height=840)
chroma=OverlayRender(NEWbase.ConvertToYV24(), OLD, debug=debug, colorAdjust=0, matrix="",  opacity=0.5, gradient=100, lumaOnly=false, upsize="Spline36Resize", downsize="Spline36Resize", mode=4, width=1920, height=840)
asNEW=luma.Overlay(chroma, mode="chroma")
asOld.Overlay(asNEW, opacity=0.5)

small=BilinearResizeMT(width/3,height/3)

Overlay(FilmGrain1(amp=4, r=1.7, seed=0, bitdepth=8), mask=small.Overlay(small.Invert(), mode="darken").ColorYUV(off_y=128, gamma_y=-128).BilinearResizeMT(width,height)).Trim(0,156731) + Trim(156732,0)

ConvertToYV12()

f3kdb(y=48, cb=48, cr=48, grainY=80, grainC=64, seed=1, dynamic_grain=true, keep_tv_range=true)

Though according to him, he never really pursued it fully to check the results all the way through.

I'm not sure that I will ever make it. It will take much computer time and my own to check the result.
Just showing that it's possible. Maybe someone else.
With different overlay modes it is possible to format output image as you want.

It is interesting seeing this came out two years before his "most complexity" merge and extraction algorithm, so I wonder what'd occur if one recycled the same script but updated the settings to make use of it.

[Stamper]

I would be in for a double barrel: First you need to auto-overlay the 2018 BR with the 2018 UHD as many of the shots are framed differently, to get the larger image height. Then you add the 2015 version grain and colors back in.

[PDB]

re: T2, my personal opinion for these sorts of regrades would be that you should always go for a print-type contrast (maybe with the highlights qualified/retained and a little more shadow detail, to be fair). I totally respect that some people like the lower contrast look they're used to, but, frankly, until home video came into being in the 80s, there wouldn't be really any reasonable expectation of seeing something with the low contrast of an IP unless you flashed the film, which would be a very specific creative choice.

There also exists a danger that an in-between compromise between this and film contrast just isn't an aesthetically pleasing look. YMMV.

Sorry that was a bit of a ramble, again this is all 100% my opinion, pls feel free to differ. In your specific case @PDB , I've greatly enjoyed your releases like Die Hard 3 that emulated a print level of contrast.

That being said, if you want to proceed with this lower amount of contrast, then I'd say you should probably neutralize the extreme highlights and shadows by desaturating them. You wouldn't really see many official home video releases where an image is completely veiled blue, green, or yellow. As would happen naturally with a print, the highlights/shadows usually are neutral to give the eye a better sense of color.

[Only exception is some early digital color grading which veered towards blanket tints, but, to be honest, this isn't as popular a practice nowadays.]

Basically the same qualms as what TAF had, which is that the saturated highlights give the regrade a bit of an odd veiled look.

What I tried in my test was simply to desaturate the image towards the light with the tonemapping (I think the yellow bit is just a trick of the eye unless Resolve does something I'm not aware of). As a result though, yes, the image does look flatter and less saturated since the brighter parts of the image are rolling off to neutral, hence my added statement that you'd have adjust the grade a bit to compensate.

As TAF already said, some added contrast would be beneficial to the image. In Resolve for this test, I essentially just bumped up the gain a bit and tweaked the tonemapping settings a smidge to get this.

Whether or not it's considered an improvement is up to you though. Given how some of the blues turn a different shade post adjustment, you'd probably have to push the colors harder depending on the scene at hand.

Message received. I just added more contrast and try somethings with the highs. I don't think I'm going to go as extreme as the DH3 project, that was an unique situation but I'll shoot for something in that neighborhood without crushing. I'll run off a couple of samples of different methods and see which people like the best.

Good luck with this project PDB, I know these films have been a passion project of yours for a while. I think I still have a test render of one of your T2 regrades

I forgot all about that project. I wonder if I still have it somewhere. Apparently I have been trying to get a T2 project off the ground for a while. 

The colours on the earlier T2 master (ie the Live or UE master) were much better than the EE master. Still not accurate to the theatrical look but not utterly borked like the EE

Are you referencing the fact that the lasers are changed to a different color in the later DVDs?

Re: T1 I can send you wav of all the files I used on my purist collection project.

Thanks Stamper

Bit of a follow-up actually, stumbled across a fairly old post Random.next made that managed to sync up these very masters it seems for at least some experimental screenshots.

I assume you've already looked at these back in the day and more recently for this project, but I thought it'd be interesting to quote the results here in case some reading this haven't / forgotten about its existence since the time it was posted...

It is interesting seeing this came out two years before his "most complexity" merge and extraction algorithm, so I wonder what'd occur if one recycled the same script but updated the settings to make use of it.

That would be pretty cool if I could get it working. I'll give it a go as a side project and see if it is feasible for me to do.