The Matrix [spoRv] *BD-25 RELEASED* — Page 10

Unfortunately you can't do that because the open matte version is slightly cropped on both left and right sides...

ColourMatch new script variations:

After a lot of testings, I managed to refine the new ColourMatch script, and ended with two variations; one with a color grading more similar to DVD, but still with some artefacts, the other with no artefacts but with a not so good (quite bad, indeed) color grading... let's call them for the moment CMnew2 and CMnew3.

I'm very happy about CMnew2, because improves the good results of the latest version; I tried to delete the artefacts and it seems it works with CMnew3, but color grading sometime is quite different from the DVD, as you can see in the following image - reason why I would NOT to use it (at least not for this project), but I added in the comparison just for knowledge:

From top to bottom:

1. CMold - the first ColourMatch script, the one I used for most part of The Matrix [spoRv]
2. CMnew1 - the new version I posted results two days ago, with less artefacts in comparison to CMold, but with a worst color grading
3. CMnew2 - improved version of CMnew1, with better color grading, even if still not on par with CMold
4. CMnew3 - no artefacts, but worst color grading than the previous one
5. DVD - used for color grading reference

So, apart the fact - now obvious - that I WILL DO the open matte version, I thought to do the following for the OAR one:

• continue to use CMold for about 90% of the movie like the old [spoRv] version, retaining 95-100% color grading accuracy
• use CMnew2 for the remaining 10%, retaining about 85-90% color grading accuracy
• for the few "survived" shots with artefacts, either using the upscaled DVD or, better, the cropped HDTV - where the side details are intact, and I could confirm that, even if in a lot of scenes there is not true, few are "pseudo-open-matte", loosing very little details at the sides, and adding many top and bottom.

It's qute difficult to say how many "survived" shots will remain, but looking at the screenshot comparisons of two days ago, where I picked up some of the worst offender, I guess they will be around 1 or 2% - a lot better than the old percentage, I think!

Plus, as I still have (of course!) the avisynth script where I used the upscaled DVD shots, I just have to replace them with the new CMnew2, make a test video, watch it CAREFULLY - another pair of eyes are welcome, though - and then locate the few shots with artefacts and replace them with HDTV or upscaled DVD (possibly using the PaNup technique and a real 35mm grain plate, to recall better the BD)... easy, no? (^^,)

Opinions?

JEDIT: I decided, ten minutes after the post, to put in action what I wrote, so now the test encoding is running, and it should be finished the day after tomorrow; then, I'll post my thoughts; meanwhile, still waiting for your opinions...

Thanks Doombot! Test encoding is finished, and I will watch it carefully tomorrow, then I'll post some thought about it.

Screenshot comparison: IIRC it's AntcuuFalb's first grain plate test, and I owe him this idea of using grain plate - thanks!

OK, I have not watched the clip yet, but just have taken a quick look at it; what I discovered is, despite the fact CMold is always very very close to the DVD color grading - sometimes 100% - that I think most of the time CMnew2 is better; still got to find out another shot with artifacts using the latter (apart the one posted before) but I'm quite sure I'll find few ones... another thing is that sometimes CMold is still better than CMnew2...

Let's see some frames together; first two without artifacts, the other three with them (NOTE: you could just go to the main comparison and switch images from there)

Frame 27784: here CMnew retain almost all the general color grading, while giving Agent Smith's face some pink, and here his shirt is white - as should be for an FBI agent, even in The Matrix... (^^,)

Frame 84000: the "infamous green beard" shot; even if in CMold Neo's face color is a bit flat, I continue to prefer it to the CMnew, which is still "too green for my taste" - even if, as I wrote before, these kind of shots are rare and almost all the time CMnew2 is way better than CMold

Frame 103953: CMnew2 has no trace of artifacts, and Neo's face is a lot better, while in CMold its color is very flat; in both the beard has equal traces of green, but not so offending as the previous frame

Frame 124008: again, CMnew2 has no artifacts, and Tank's face has better depth, while in CMold its color is quite flat

Frame 155508: some colors are lighter than CMold (and consequentely than DVD), but overall grading is right, and IMHO better - indeed water seems fresh here, while stale in CMold...

Conclusion: I think I will stick with the CMnew2 color grading for the whole movie, replacing the few shots where clearly CMold is superior (that frankly seems to sparsely appear), and eventual shots plagued with artifacts will be replaced by upscaled DVD or cropped HDTV; a test version should be encoded to support (or eventually confute) my theory.

35mm footage won't be too far away now, 6-8 weeks hopefully.

The print is pretty scratched up, mostly useful as a colour reference to correct the BD to the theatrical release, but still fun to watch.

I think many people will be surprised that their memory of the film doesn't match what is actually was in the cinema, it had a very blown-out, almost bleach bypass look in many scenes, with white burned out skies and a high contrast look.

Looking at the scenes, correcting the BD wouldn't be too hard to match it to the print.

LAST UPDATE:

After A LOT of technical problems (that I spare you to know, it's so boring...), I finally finished to encode the movie using my ColourMatch2new script; it worked pretty well, as wrote previously; in the few frames with artifacts (still to check the whole movie, though) I managed to make some further corrections, but...

The problem is, x264 crashes after few hundred frames, maybe due to the high memory and CPU requests asked by the complicated AviSynth script, and the low power of my system... nevertheless, shots with artifacts should be few and shorts, so I thought to use this last tweak sparingly, and here you are the result on the most offending frame:

Not perfectly regraded as the DVD, but still better than the original BD, color wise, and so all the movie will be sourced from the BD, avoiding the use of HDTV or upscaled DVD shots; I also decided to use the old ColourMatch script for very few shots, like the "green bearded Neo", where the new script clearly can't beat the old one.

So, I'm near to the V2.0 - until I will receive the laserdisc rip of the movie; then, I should redo everything from scratch! (^^,)

drngr said:

If the problem is memory consumption, you should be able to just encode a lossless intermediate file.

I did it... it's the same! Well, I will encode the small chuncks separately as lossless files (like the main one); then, I'll join'em all to produce the final x264 encoding.

Any new progress?  (Yeah, South Park made me think to look up this thread.)

Also are there any other screenshots than the one in post 237?  I'm not convinced the CM2new tweak look much different than the BD screenshot.

I'm pretty busy these weeks, but I'll do it, sooner or (much more probable) later; here you are somenew screenshots using the same script as the one in post #237 (CM2NEW tweak... too long name, so
I changed it to CM2x) - frankly, even if not perfect, it works much better than the old ones, and, despite
the fact that some parts still should use another source (HDTV probably, or upscaled DVD) - like the now
infamous "green beard Neo" shots - these parts will be a lot less than the ones I used for the first version...
and, frankly, I don't think they look like the BD, but at the contrary I think they are closest to the DVD color
grading than the BD. What do you all think?

There's a 35mm transfer in the works?  Does that have a separate thread?

I'm using the DVD as color reference (the HDTV has the same color grading of the DVD, by the way), but I would be happy to see if the laserdisc is different - but I'm quite sure it will be the same.

And, thanks for your kind words - I'm alive and kicking... not kicking a lot, actually, but "I'LL BE BACK!" (^^,)

I've seen some Laserdisc caps on the LDDB forum and it's the same master as the 99 DVD. The 35mm print has a totally different color timing though.

So, will be better to use the 35mm scan as color reference, instead of DVD?!?