For color correction, would it be possible to have a script generate a LUT using the 4k77 print as color reference for each shot?
The last reel is finishing up now. I’ll be transferring the files off (that alone takes 30 odd hours) and will be creating a ProRes proxy over the weekend.
I will have the version ready for checking early next week, thanks to the people who contacted me via PM, I will break up the reels so that you have one reel each to check. Please use the burned-in timecode to report any issues.
Thanks again for the help, I’d like to ensure we find any issues while the print is still available for re-scanning if needed.
For the Pro-Res, how will you be encoding that? 12bit CINEON or another log curve?
I’ve been working on a restoration of sorts for the film Koyaanisqatsi; in my view, the “remastered” soundtrack they used is in many regards not up to par at all.
Through a few sources, such as the laserdisc PCM* and the original album release, that I consider to be the best quality, I’ve been able to put together a “restored” version that I think sounds much improved.
Official soundtrack mix: https://www.youtube.com/watch?v=eNNJrHR53ZA
Remastered soundtrack mix: https://youtu.be/q86RvSxAy04
This will be released in 5 channel and stereo mixes.
*The laser disc PCM will also be released as a preservation. I performed a complete restoration on it, removing clicks, pops, dropouts, line hum, and reduced hiss and noise.
I put together a screencap gallery for those who want to see some stills from the 4K version:
The highlights aren’t blown out on the UHD; they just look blown out because of the tone mapping. The only way you can tell if highlights on a UHD are blown is if you look at the image before the EOTF is applied. Otherwise, it’s likely that the tone mapping clipped or compressed the highlights because your display cannot get as bright as the Dolby Pulsar it was graded on.
This is one reason why Dolby Vision (and maybe HDR10+) are coming on newer releases; it has additional meta data to allow the colourist to specify how the image is to be presented on displays that cannot reach the brightness necessary to properly display the picture.
Are there any copies of this cut floating around? If so, I’d love to try and insert them in to the 251 minute Blu-Ray!
Anyone know who produced this release? Would love to get in touch with them to work on a 4k release with a 10bit DCI /P3 gamut.
Reading this thread makes me realize why I greatly curtailed my participation in famdoms years ago.
https://youtu.be/u8FgrPpuuRk <- Here’s a rough edit of the pod ride/Enterprise tour (includes a new composited shot). Let me know what you think! Need 3rd party perspective on how it flows.
If anyone has a bit perfect copy of the PCM track from the TMP LD please get in touch; I found out the copy I have (which was converted to 48Khz and synced to the Blu) has a LOT of clipping in the latter half of the film.
@Williarob Wow thanks for the information! I’ll definitely check out the videos and try to get that batch processor running. Your suggestion concerning AfterFX is similar to what I use within Nuke; I set up the two sources, insert a grid align on the HDTV source, merge the two sources via a Plate_Align gizmo (it works a LOT better than difference mode to my eye), and tweak the grid points until it gets close enough.
Processing via photoshop would at least result in me having to manually correct my grid transform node less frequently.
Oi, all this work to fix some idiocy on the part of a colorist.
@ElectricTriangle: I might consider doing that just to make the process easier. However, using Photoshop or Nuke’s autoalign doesn’t get me all the way there because it doesn’t correct the non linear distortion within the frame.
Finished restoring the Wormhole exterior shots via combining the complimentary information in the Blu and HDTV copies. Took several hours… and I still have to dust bust the footage! Amazingly tedious. Basically have to:
-isolate the footage in Avid and export every shot (for both the HDTV and BLU transfers)
-generate a color correction LUT via the Color Matching Tool
-throw each shot in Nuke wherein I need to:
—manually align them with grid warp (the frame geometry differences between the transfers is variable)
—apply the LUT
—set the luma key so I find the right transition points between the footage
Hoping that my method can be streamlined somewhat. I don’t suppose any of you know how do setup image registration software like Elastix? Or perhaps a method using Avisynth? The manual alignment of the footage is the most time consuming part and if that could be automated to some degree, that would be amazing. Also automating the color matching, like say, instruct the program to color match & generate LUTS for all PNG pairs (ex 1a.png with 1b.png, 2a.png & 2b.png, etc) in a directory would also speed things along considerably.
I’ve been tinkering with the idea of adding some of the details of Andrew Probert’s “cargo bay” matte painting to the version Matthew Yuricich completed for the film; no offense to the master that was Yurichich, but the rendition of the cargo-bay in the final film is very boring to look at.
Still needs to be grained and color graded. Thoughts?
The behind the scenes documentary on the bluray shows the transfer before the DNR and contrast boost, although it still unfortunately doesn’t include any full shots.
I believe that footage came from the same scan that was used for the HDTV master; the frame geometry is different from the Blu-Ray.
Hah it will be even more magical if I can find a better way to align the transfers; right now I’m doing it manually and, while it works okay, it’s tedious and the alignment quickly changes within the sequence (notice how the right bottom corner has some ghosting due to misalignment).
Back from the dead with some new power-tools!
I’d like to present a frame from a restored version of the E reveal. The Blu-Ray transfer is missing the dark and highlighted areas of the image, and I was able to restore those using the HDTV transfer, along with the film grain that was removed from the Blu-Ray. Best the film has ever looked in my humble opinion!
I’m also working on dust-busting the deleted scenes and concocted a superb upscaling script so they’ll feel less out of place amongst the HD footage. If any of you want to do Director’s Cut or TV Cut restorations of the film in HD get in touch and I’ll find a way to get the material to you.
Maths major here. I understand the concepts at work in regards to removing grain. Some detail is lost when grain is removed… it’s just whether or not it’s something that can be perceived by the human visual system. Removing most of the grain is a very common practice in post houses when it comes to grading film footage; as long as it’s done correctly and the grain is reinserted after grading is completed, the viewer won’t notice a reduction of detail. This is all beside the fact that what we perceive to be detail is contingent on many factors beyond the presence of actual detail.
Anyway, as mentioned, according to what I see on caps-a-holic, most of the blu-ray shots aren’t lacking in detail insomuch as they are lacking in grain and contrast; in the shot where Indy is holding an anachronistic rocket launcher, you can easily discern a detail as fine as the fringes of the flags in both transfers. Also notice that, in the HDTV cap, the flags have a very harsh border in the form of a black outline.
Also worth considering that the limited bit-rate of the HDTV capture will result in loss of detail to a further extent than the Blu (especially if some of those bits have to go to representing the grain).
Out of curiosity, are any of you using monitors calibrated via 3DLUT in a light controlled room? If so, what are you monitoring your video feeds with and how do you know what you’re looking at is accurate?
Also, with regards to the Blu vs the HDTV cap: from what I can see on caps-a-holic, it doesn’t seem that the Blu-Ray capture is generally lacking in detail (barring some shots) insomuch as it simply has had the grain removed and the contrast lessened (contrast effects perceived “detail”). Additionally, the HDTV capture’s grain is larger and has horrid compression artifacts. I’d wager that, if the blu was re-grained and regraded properly, it’d look better than the HDTV capture or print-scan.
Yep, I just clicked-through his 1hr second video and he never compares how they look in motion, all he compares is how they look zoomed in on a screenshot.
??? He was arguing against the merits of a specific trend in the industry… basically the spin by equipment manufacturers that higher pixel-count pipelines equates to a clearer picture in a way that is noticeable to an audience. Motion resolution is a completely different subject.
Steve never mentions the problems of aliasing or other strobing artefacts caused by lower digital resolutions,
Why should he? He isn’t arguing for SD capture/presentation or any nonsense like that.
He also doesn’t specify what film stocks he used… Especially given that he had 4 different types of digital sensor, but only one type of 35mm film.
He mentioned the IMAX and 35MM being the same stock. In any case, I’m quite content to trust that this guy, who…
-has shot 13 features on film,
-is been somewhat nerdy about his tools,
-is part of the AMPAS Science and Technology Council,
-had their conclusions endorsed by the president of the ASC,
-is lensing the new SW film,
… probably knows enough about what he’s doing such that the question of “if he used the right film stock/scanned it correctly” isn’t one worth asking.
This is actually not the case. I have a 2k copy (2048x858) of a 35mm scan, which shows the pattern fairly clearly.
Was the scan performed at that resolution or was it done at a higher res then laster down-sampled to 2k?
Fair enough. Quick note though: the ITU paper said that “The highest resolution that the expert assessors could discern in the highest performing movie theater was about 875 [lines]” with the qualification that “a wide range of resolution values will be obtained, depending on the stocks, laboratory, type of printer, and so forth.”
Also: it’s important to clarify that just because you need to scan at 4k+ resolution to extract detail from a print doesn’t imply that said detail can only be resolved at that resolution in a digital master; as I mentioned previously, this seems showcase the necessity of oversampling vs the resolving ability of the film.
Anyway, I digress. Scan looks lovely.
The pdf I linked to goes in to enough depth such that it might help you to answer your question; tests were conducted in a theater, using (for the time anyway) top of the line film stocks, properly set up projection, lenses, etc. Even then, the apparent resolution of a release print to the audience member was stunningly low.
I’m not trying to cast doubt on your expertise here, and I’ll agree that it’s important to scan 35mm at 4k+ for full detail retrieval due to anti-aliasing and the like, but if DP Steve Yedlin can’t get fresh 6k scanned 35mm filmstock to resolve more than a 3k Alexa in a controlled environment, there is no way a release print suffering several generations of degradation is going to outdo that (especially when created via older processes).
Of course these are generalities and the resolving capability of film isn’t a very linear process (it isn’t with digital cameras either).
This is also a good read as well: http://www.motionfx.gr/files/35mm_resolution_english.pdf
EDIT Yes I can see what you mean about the ribbing. That being said, it’s important to remember that just because you need to scan at high resolutions for details to become visible doesn’t mean those details “exist” at that resolution/couldn’t be adequately resolved at a lower resolution (the reason why probably has to do with Nyquist–Shannon).
3500 lines aren’t enough? 😉
The final scans will be in ACES colourspace.
Nice; hopefully we can get a P3 gamut version produced.
Haha I was referring to how the apparent resolution of release prints is rather low, which I guess shouldn’t be too surprising given that 35mm negative struggles to resolve as well as a Arri 3K camera. Even moreso the case for older theatrical prints; I’d imagine newer print making processes introduce less degradation vs older methods.
For those who’d like to go down the rabbit hole, the director of photography for The Last Jedi made a very interesting video lecture about how different different cameras and formats influence the apparent resolution of an image… basically arguing that the focus on pixel count is misguided; his comparison showing that a 11k scanned 70MM Imax negative struggles to resolve as clear as a 4k master from an Arri Alexa 65 is rather sobering.
it should resolve well above 2K.
I dunno about that; if a 6k scanned 35mm negative has trouble resolving as well as the Arri 3k camera, and take prob 20% off that figure for generational loss (in reality I think it’d be a lot more), you’re looking at about 2.4k.
Here’s some further reading on the matter that’s rather interesting;: http://www.reduser.net/forum/showthread.php?59036-Resolution-of-35mm-film-and-70mm-IMAX-film