little-endian

imsorrydave2448 said:

Some directors have intentionally increased the effect of the film grain during scanning. Stanley Kubrick famously did this with Eyes Wide Shut.

If so, then entirely legit and shall be preserved, bar none.

And if it’s part of the technology, it’s intended.

Not necessarily when considering the entire history and not just today. Sure, if done so today, it most likely is an artistic choice. But back in the days, many flaws were accepted because it simply wasn’t possible in a better way.

And nowadays, you have to intentionally choose to use film.

I see that you’re arguing from today’s point of view. Just to halfway stay on topic, that is already questionable in the case of “The Mask” as video cameras were still quite flawed at that time compared to film cameras. So one may raise the question how much here was really an intentional choice or simply dictated by circumstances.

As for 24fps, pretty much every person who’s not Peter Jackson, Ang Lee and James Cameron knows why making every single movie ever into a higher frame rate movie isn’t going to work.

Which can only be a subjective reaction as technically, taken details such as the different limit for exposure times out of the equation, the higher the frame rate, the better the result will be in the sense of higher time resolution (higher nyquist frequency before aliasing occurs). Even for the subjective part, one shall differentiate between the unchangeable preference of people, which also is entirely legit, and the simple lack of being used to it (which I conjecture, is by far the main reason). With film, people simply expect a stuttery reconstruction although higher frame or field rates aren’t anything new if one considers PAL/NTSC standards since decades.

Either the amount of movies released is going to have to be cut drastically(studios would not be able to do this), or the quality is going to have to drop significantly. And the latter would defeat the purpose of doing a higher frame rate.

Well, that is something concerning company politics, willingness for change, budgets, etc.

My approach here was purely from a technical / information theoretical viewpoint. Hence we certainly can agree on that The Mask shall be decently scanned with, please, as little “adjustments” as possible and definitely no stupid grain filtering.

rwzmjl said: […]believe it or not, that’s an intended part of the image![…]

I read that argument often lately, but the word “intended” is a crucial premise here as in many cases, it isn’t really but rather a side-effect of technical limitations (which includes our visual system introducing retinal noise as well). Same goes for the stone age low rate of 24 fps only, causing tremendous amounts of temporal aliasing and making a subjectively stutter-free reconstruction difficult.

So while I can understand the sentiment for artistic reasons and preservation, technically, noise isn’t part of the original image which one tried to capture. Noise also limits — or rather defines — measurements such as the SNR.

I’m also entirely up to “preserve” it in terms of not filtering it out, as not only it isn’t possible to do so without losing information from the actual image anyway, but also because it may fit a certain desired look. However, so do noise and crackles on vinyl records which may be preferred emotionally, but shouldn’t be rationally.

One technical advantage of noise in the source shall not go unmentioned though: (self)dithering, enabling a theoretically smooth reconstruction of an unlimited number of shades (and not just 50 of grey) without ugly banding. However, dithering can be artificially added later during the A/D conversion as well in an mathematically optimised way even so even that is a rather far fetched argument.

BoingoBanshee said: Actually, this IS real 4K. 4K isn’t strictly “the camera negative”; you can do 4K scans for any film element.

Well, the same way one can record any audio material at any given sample rate of choice, but that doesn’t mean the original carries that resolution in the first place.

For example, I’m well aware of those “Fake 4K or not” websites where, in the case of, let’s say, a 2K digital intermediate, they rightly criticize that no “Real 4K” performance may be achieved and thus the 4K (actually “UHD”) release is “fake.” However, that view is also a bit misleading and incomplete, as the assumption that film material automatically delivers 4K or whatever resolution just because it was scanned as such technically isn’t correct either.

In practice, some 2K-only productions (at least in parts, such as “Collateral”) may be categorized as “Fake 4K,” whereas some fuzzy “as good as it gets” film source, scanned in 4K, is called “Real 4K,” despite the fact that it may never reach that resolution either and never has. Many may be above 2K level, no question, just to make a point here.

One would actually have to measure the real spatial resolution of such sources (in MHz), and then one could derive the required or equivalent resolution in “pixels” (which theoretically is twice the analog bandwidth according to Nyquist/Shannon) to preserve the original without quality loss.

Back to “The Mask” - I admittedly have no idea what spatial resolution your exemplar comes with, and a 4K/UHD (and maybe even HDR) scan certainly won’t harm. So excitement is justified as the official BD release definitely is mediocre at best by today’s standards.

blakninja said:

This won’t be a real 4K as we are not scanning from the original camera negatives.

Yep, and even then, it isn’t guaranteed that what has been originally shot back in the days reaches the spatial resolution equivalent.

Scanning at 2K would be borderline and we might miss some details[…]

Most probably exceeding the official Blu-ray release though which certainly doesn’t reach decent HD level (which is the whole point of the project besides the open matte aspect so to say of course). Guess they took some mediocre HD master which was good for the DVD release and reused that. However, at least they refrained from ruining it with orange & teal color grading, heavy DNR and other sins.

[…] and scanning 8K would be overkill and we just get a bigger image with not necessarily more details.

Definitely. 96kHz/24Bit vinyl rips are greeting as analogy.

It will definitely have more resolution (details) than HD (1080p), at least for the non SFX scenes.

If so, even better and more exciting.

On top of this, the grading should be closer to what you’d see in theatres (not that the Blu-ray grading was too much messed with) and we get to see it in open matte.

Great indeed. And I guess you intend to also refrain from any filtering, DNR, sharpening or other fuss, right?

MrQuiche said:

Definetly interested too, i’ll wait for some preview pictures too but i’d love seeing this movie in 4K. Thank you in advance 😃

No intention of spoiling this great project, but you won’t see it “in 4K” with this either, as the film print will be far from that resolution level. It’s already questionable whether 2K/HD will be reached, as it’s usually a copy of a copy of a copy (which reminds me of “Fight Club,” which could also benefit from a long-overdue UHD release, but I’m drifting …).

Thanks a lot, MJBenito. Yes, I’ve noticed that your nice project video file doesn’t contain the DTS version so I’m happy to contribute here. I also have the CinemaDTS (and the LD audio as mentioned before) at hand, but I yet again will have to investigate how that is properly decoded now level-wise as there used to be quite some confusion about that (at FanRes), especially regarding the level of the part below 80Hz which is embedded into the surround channels as that format came without a LFE back in the days.

Anyway, on a quick run, I want to spend a few words on The Mask and hope that you don’t mind a critic review of your upscale/remaster. All that is meant to be constructive feedback.

It is clear that the used source in the unfortunate lack of a fresh 4K scan is pretty dated and most probably comes from a time where it was just “good enough” for a SD release on DVD. As the comparison at caps-a-holic shows, the BD release is barely any sharper that the DVD, but at least they didn’t fuck it up (yes, James Cameron, I mean you!):

https://caps-a-holic.com/c.php?go=1&a=0&d1=3438&d2=3437&s1=31680&s2=31672&i=0&l=1

Now your release, and this is entirely my honest, humble and at the end subjective opinion, introduces a different kind of issues, similar to Titanic: What at first looks sharper, isn’t really and if something may irritate on your releases, despite the way better (in terms of more neutral) colors without any doubt, then it’s that typical “oh, quite sharp … but where are the details?” kind of impression. Technically this makes perfect sense as one cannot create information out of nowhere (overly hyped AI-processing nowadays or not).

Let me undergird the argument with some screenshots:

https://ibb.co/XW0VqtN
https://ibb.co/7493nmT

Yes, your version appears sharper at first, but when paying attention to the pedestrians, they really start to look very artificial. Not that they looked great to begin with, but not as waxy and “Playmobil”-manikin-like.

In this shot, when looking at the grill/ventilation part to his right, one can witness a small (!) amount of lost detail on your release which is still present in the original (which would also support my observations with Titanic). So while your used AI program adds (perceived) sharpness by boosting the contrast and other techniques (without the usual sharpening artefacts, I have to say, no ringing, no halos), it slightly (!) reduces the spatial resolution of the original as well:

https://ibb.co/9nRM8Fw
https://ibb.co/94YN6Mk

I hate to say it, but for me, that without any doubt too fuzzy original BD has a more consistent look despite its many flaws. On your release, although I want to love it for the effort alone, I cannot get rid of the feeling that something is wrong as one has the conflicting impression of sharpness and at the same time lack of real details which the added noise (for me at least) can’t really mask (no pun intended here).

If not too much trouble, as with Titanic, I would also appreciate a version which is only color adjusted and nothing else. However, it remains to be wished that “The Mask” finally gets a decent UHD release as the not so great BD even for its time, doesn’t give that still nice movie its justice.

Thanks for your reply, MJBenito.

MJBenito said: As mentioned in the original post, the 3D bluray is the source used for this 4K Filmized remaster.

Yes, that was my understanding, however thorough comparison of that original and your version raised doubts about that so thanks for the confirmation.

MJBenito said: I used my own Topaz Video AI settings after many tests to get the most natural look. That means that compression has been fixed and details may have a been recovered during the process. I made sure that no sharpening was added in the process because it was the main goal of the project : get rid of the oversharpened look of the official 4K remaster.

While entirely appreciating your great project and also approach when it comes to the color grading, from a critical point of view, one has to state that by trying to recover details, some are also lost (minor, but still). Also, as my screenshots show, in fact there is artificial sharpening in your release which isn’t present in the original, a fact that actually led to doubting which source you had used again.

Thus, if your time allows it, I personally would highly welcome a version with only color adjustment applied but nothing else. No upscale, no AI processing. Same goes with “The Mask” where I intend to comment on separately in the respective thread.

MJBenito said:

Project is ready guys.

In terms of you’re finding the time to start it or that you’ve already finished it?

Also, how about the inclusion of the CinemaDTS as I suggested?

@MJBenito: if you find the time, it would be great if you could answer the questions I raised about the source and the sharpening differences, thanks.

After I had mentioned a “V3” to be even more preferable (combining the 2D version which has less DNR to begin with and the open matte 3D one), I was contacted by a member who suggested some Indian release of Titanic which was supposed to have no DNR and being open matte as well. Once I got a copy of that, it quickly turned out however to be exactly the same H.264 stream (just on one disc instead of two) from the left eye of the 3D version. Same size, same artefacts.

Taken the messed up UHD BD and older DVD or LD releases aside, there seem to be only two official HD variants: the less noise filtered cinema scope one provided by the US 2D BD, Spanish 2D BD, etc. and the more noise filtered open matte one provided by the US 3D BD, Indian 2D BD and some scene releases like the usual suspects such as “KRaLiMaRKo”, etc.

When I compared all that, I’ve noticed differences in sharpness though between the official open matte and Benito’s regrade.

In this scene, the carpet appears significantly sharper:

https://i.ibb.co/MVHSsGT/Titanic-Indian-2-D-OM-US-3-D-OM.png
https://i.ibb.co/R4FDGS8/Titanic-MJBenito-Regarde-V2-OM-2.png

Here Rose’ chain also looks a lot sharper. At the same time, some original detail is lost though, apparent on her dress as I mentioned before, but not only compared to the cropped 2D version, but also between the 3D original and Benito’s regrade:

https://i.ibb.co/y60v2mv/Caledon-Rose-Original.png
https://i.ibb.co/5GKc8hk/Caledon-Rose-Benito-Regrade.png

A zoomed part of her dress to emphasize the difference; parts of the dress’ strikes are masked by the noise:

https://i.ibb.co/SnSzsqV/Rose-Dress-Original.png
https://i.ibb.co/dDhXmtQ/Rose-Dress-Benito-Regrade.png

Which raises the following questions for me:

What’s the source? Is it indeed the official 3D one? Then the subjective sharpness increase is impressive but would also mean that you applied artificial sharpening and not only upscaling.

Overall, comparing the two, the data reduction, JPEG-like artefacts are somehow entirely replaced by some “error diffusion”-like dither. If my assumption is correct, I would be interested in a V3 with only upscaling and color correction and no filtering as at some point the detail of the dress has to go lost.

But maybe you had a different source after all and I happen to have a 3D BD which coincidently shares the “magic” Indian source.

In short, I have been successfully confused.

@davicroquette, @justArobo

PMs sent.

I’d also be interested in that and could not only throw the DTS-rip from the US-Laserdisc onto the pile for that special release, but even the Cinema DTS one. Doesn’t get more truthful than that. 😃

However, both would have to be synced to the video and in case of the Cinema DTS, it would be technically (not necessarily audibly) be better to reflag the BD video stream which most probably comes in 24/1.001 fps to 24 fps so one doesn’t have to resample the audio at all and maybe just use that or make several variants.

PMs sent. If anyone else hasn’t gotten it yet, let me know.

Hold your horses, rescue is coming very soon …

waffle said:

I recommend using MPC-BE

Thanks for the hint, I’ll check it as VLC also seems to be terrible at displaying Dolby Vision content properly as well (totally wrong colorspace mapping).

MJBenito said:

Can you also share it with me ?

Sure, just give me a poke once returned and I’ll send you links.

Now that Cameron missed the mark on the new 4K transfer[…]

A nice way of saying that he and his crew idiotically messed things up after they had about three damn decades to get things right. Just pathetic.

Theatrical trailer 35mm 4k Scan (TOP) Theatrical release print scan v1.0 (Bottom)

Judging by the screenshots though, the one above seems to have even more black crush than the original scan which already has severe loss of details compared to the video variants.

We’d need the negative as cinema copies probably all have low shadow detail which overall makes me think that the HDTV open matte and the D-Theater one remain being the best versions available.

While the whites appear a hint more neutral to me in V1, at least on my computer screen via VLC (will check in on the A95L later on), the noise distribution I find nicer on V2 as it masks the fuzziness and better takes away the “wax effect” caused by the filtering in the original. An interesting psychological effect that it appears sharper although there isn’t more detail (sharp [no pun intended] tongues might say that this is part of film material’s secret).

For the others, @19:02, look at the guy’s above:

https://ibb.co/YXwLZ5q
https://ibb.co/176Btz2

Update after checking the differences on the TV:

Interestingly, the color shift is way less apparent than on the PC with VLC. As for banding, I still have some in the V2 (especially in the underwater sequence at the beginning), but I suppose that this is already embedded in the 3D source.

About the HDTV open matte rip:

jakeandelwooduk was so kind to provide me the HDTV open matte rip (in 25 fps) and again, it is quite different from the 3D one. A lot more color saturation and the framing also seems to be slightly different (besides the lower bitrate and thus data reduction artefacts of course):

https://ibb.co/PxfG3Bq
https://ibb.co/7jdh9pM

@MJBenito:

While I was actually looking for differences in regard to banding (where V2 is better indeed), I’ve also noticed a color difference between V1 and V2 with the latter tending to be slightly greenish now compared to V1:

https://ibb.co/fksjxtn
https://ibb.co/2kNd0HN

Is that intended?

MJBenito said:

I’m working on Premiere Pro with Lumetri color adjustments[…]

Thanks a lot for your insights. I want to start doing that as well, as the silly color grading of the UHD BD of “Training Day” (otherwise a welcome improvement over the BD) similarly annoys me, asking to be “fixed”.

https://caps-a-holic.com/c.php?go=1&a=0&d1=18089&d2=18105&s1=207597&s2=207955&i=0&l=0

In this case, considering the Open Matte version is sourced from a h264 4:2:0 8 bits encoding, I thought the whole 4K Filmized Remaster process would have worked in the same color space… Only to find that the final image had banding and macro blocking issues.

Yes, the reason for that most probably is that during the conversation, not only get the color values shifted around, multipied, divided, etc., but furthermore in the more versatile floating point domain and then transfered back to integers, necessarily getting rounded to the 8 bit (or rather ~ 7.78 bit equivalent, as video signals often only use the values 16-235, so called “limited range”) which then results in distortion; at 8 bit per color perceivable as banding.

My guess would be that for some reason, as so often, no dithering is applied which at 8 bit is a must to avoid banding artifacts. For instance, many Blu-rays also have lots of banding embedded which would not be necessary at any bit depth if one accepts the higher noise floor in return.

That’s why I made a V2 from scratch using 4:2:0 10 bits for the whole remastering process. You’ll surely notice differences from V1 to V2 during deep ocean scenes and final scenes of the movie.

Good move. While dithering is desirable even at 10 bit, probably it isn’t noticeable anymore and when dealing with lossy codecs, higher bit depths might be preferable to dithered 8 bit due to the additional noise it creates. On the other hand, you added some to simulate the film grain anyway. Interesting that it apparently didn’t act as dithering itself to prevent the banding here.

Of course I’ve seen the post regarding the HDTV version but I wanted the best encoding available for the open matte. Even if that means a little more DNR baked in the source.

I contacted him as it would be interesting how that version looks but most probably yours will win overall by cleaner encoding source alone, yes.

Superb work of yours anyway and the fresh/neutral colors are a pleasure compared to all the official releases.

MJBenito said:

Hi, I’m quoting your last post but I’ll try to answer every question you’ve asked in this thread.

Thanks. But let me pick up those which got unnoticed:

• What tool(s) do you use for that kind of color adjustments?
• Do they offer any option to apply dither the output before the encoding stage?

BUT it is the only version available in 1.78.

Except for that HDTV version which jakeandelwooduk mentioned, isn’t it?

I tried to compare it with the scope version and besides the fact that not all shots are open matte but sometimes slightly zoomed, the geometry of the shots are also not always matching. After thoughts, I decided it would be too much work to blend 2D and 3D images together cause it would have to be done shot by shot.

I concur, same diagnose. I supposed the effort is also tremendous with “only” the color correction, isn’t it?

But my work on Titanic will stop with the 4K Filmized Open Matte Remaster.

Fair enough.

And the whole project that will also get the 4K Filmized Remaster treatment with adjusted color grading.

What would be great!

V2 is now available and should sync with you all.

Many thanks for sharing and all your great work. Will be enjoyed, backed up and cherished.

Again I wondered what could be accomplished if anyone sane would get access to the original negative and scan that with only minimum color adjustments so it looks right on today’s screens with pristine HDR, preserved film grain, open matte, etc.

Sigh …

A few further thoughts on that project which might call for a “V3”:

While MJBenito’s color adjustments are great and the slight noise layer maybe rescue at least some filmic style, it cannot hide the already “scrubbed”, de-noised and somewhat waxy nature of the source.

Also, directly comparing the 2D Blu-ray release with the upscaled 3D one, the former still has a bit more resolution (visible on the clothes) so I wonder whether an open matte version could be created out of the 2D source with its higher quality and the missing information above and below stitched together from the 3D open matte version. However, the image angles seem to be a bit different and I’m not sure if either eye frame (left & right) from the 3D BD matches the original 2D film frames.

https://ibb.co/VCCCG5C
https://ibb.co/sPwHC6p

There is also an extended cut floating around from 35mm where the additional scenes don’t seem to be sourced from the film, but rather some Web-DL streaming video.

Sigh, all in all, it is so damn frustrating considering that those morons are sitting on the best source possible, the negatives, and apparently can’t simply release that stuff without screwing things up.

I can’t decide what I find more retarded: the stupid AI sharpening on the UHD-BD, the goddamn orange & teal ugly color gradings on both the Blu-rays or the filtering of the 3D-Blu-ray (the latter one maybe being the least non-sensical in regard to the noise which might distract in a 3D experience, although I doubt even that as our human vision isn’t noise-free either thanks to retinal-noise which can be very visible as well). Even if, there is no good reason not to release the open matte in 2D without the damn noise filtering.

So at the end, neither version is really great.

@MJBenito: maybe you could consider to release also a regraded version like this one here, but based on the 2D cinema scope source with its better quality or test another version with even more noise added as the Blu-ray has a lot more (color) noise than its 3D counterpart:

https://ibb.co/C8tfQ09
https://ibb.co/W6kqXKp

That’s great news. Thanks in advance from my side for all the effort!

Some remarks about banding: While I’m not too familiar as of now when it comes to video encodings, in the audio domain, it is always (!) advisable to use dithering in order to transform distortion (in the video domain that corresponds to the infamous ugly banding) into an equally distributed error, resulting in way more pleasant (and often enough not even noticeable) noise.

With video, that may raise the data rate requirements in conjunction with lossy codecs - which would also be my guess why dither is apparently rarely used (besides pure ignorance I suppose) as technically, no bit depth has banding as a direct consequence, but only determines the noise floor under ideal circumstances. Even with 10 bit or whatever bit depth, dither should be applied in theory, but I suppose it is much less of a problem with > 8 bit / color, just as it is neglectable when dealing with 24 bit PCM audio.

So in case you encoding software actually supports (random) dither like the renderer “madVR” does, maybe you can enable that for the next release.

@MJBenito

Having ranted in several forums including Youtube about that insane UHD release of this - from today’s perspective - real classic, I’d love to get a download link as well and highly appreciate the chance to have this movie in a less flawed variant so to say.

Besides the hip AI sharpening, it is beyond me, what moron invented that shitty orange & teal - based kind of color grading in the first place, spreading like a decease. It is so incomprehensibly ugly, unnatural and annoying, grrr. Why can’t they just scan the damn thing and refrain from fiddling around with it, is it so difficult not to mess it up?

Out of interest: with what tool(s) did you perform the color adjustments?

Since the open matte 3D version of the Blu-ray is still more processed than the 2D one, I wonder whether it would be possible to take that as the source instead and extend the missing frame parts with the ones from the 3D version.

sidshady12 said:

All I care about is maintaining the accurate sound but presenting it in a 5.1 format, as this film makes heavy use of surround sound.

Then eac3to is probably the tool you wanna use. Just call it with the -down6 parameter and it will mix down any 7.1 or 6.1 source to 5.1.

johnkimble said:

one flew over the cuckoo’s nest - another point less 5.1 mix w/ added foley.

Well, besides that any LFE is pure legacy to begin with and wouldn’t really be required in any PCM form, in case of such movies, that “.1” doesn’t make any sense at all of course. Given the dialog-heavy genre, here it boils down to being essentially a 3.0 mix disguised as marketing-friendly 5.1 as usual.

Anyways, since I recently acquired the LaserDisc mono track from a friendly member at fanres.com, I have to say that I actually prefer the remix to that one as the LD’s audio for some reason has a noticeable metallic/tinny/roary sound in the background in several scenes which I personally find awful. I have no idea what causes this as I’ve already noticed that in other, way more modern releases from time to time. Sounds a bit like the effect called “sleepy hollow”, like being played through an echoing metallic tube (also the newer mix of “Léon - the professional” has it as I reported here).

Apparent scenes are:

08:57 (McMurphy having a seat in the doctor’s office)
59:10 where they are on the boat

In both cases, the LD sounds very tinny with the sounds of the waves in the boat scene barely hearable whereas none of this happens with the 5.1 AC3 provided by the Blu-ray.

Since I cannot say for sure whether it is just the way it was originally recorded (and magically removed later on) or something which went wrong with the contact’s capturing, any other of your experiences to be shared would be appreciated.