wildlava

The “23.976 fps” frame rate is a direct result of video frame rates established when color TV was developed. Doing the 3:2 pulldown process (used to convert film to video), and matching the NTSC frame rate of 29.97 fps, we get:

24 * 29.97 / 30 = 23.976 fps

This rate did not result, accidentally, from the imprecision of old mechanical film cameras/projectors. You are right that 24 fps was chosen for film, but, as you say, the old cameras did not have the kind of precision to all be “actually shooting at 23.976.” It was a result of the chosen NTSC TV scan rate. So my point is that old film cameras shot at 24.0 fps (or rather as close to that as mechanically possible: +/- some error), not precisely 23.976 (which is used for video and derived from the formula above).

Darth Lucas said:
Your understanding, as well as your assumption, are incorrect.
Very few movies, unless they are very early digital movies, were shot in true 24fps.

My understanding is that 24 fps (true 24.0 fps) was settled on in the early 1900s (just Google it). This was for films (not video). And that makes sense: why would someone decide on a film frame rate of “23.976 fps” back in those days?

The 23.976 fps rate is related to the 29.97 fps of color TV. 29.97 was used to avoid problems with old TV hardware (see https://en.wikipedia.org/wiki/NTSC).

The Wikipedia page for “24p” (in the “23.976p” section here: https://en.wikipedia.org/wiki/24p#23.976p) states, “Nevertheless, even in NTSC regions, film productions are often shot at exactly 24 frame/s.” It also states that, “Some 24p productions, especially those made only for NTSC TV and video distribution (e.g. in Canada or the USA), actually have a frame rate of 24 * 29.97 / 30 frame/s, or 23.976frame/s.”

So my understanding/assumption that most film productions destined for the cinema were shot at true 24.0 fps is pretty reasonable (if you find a reference that contradicts this, please post it). And if true, the Star Wars trilogy was also shot at 24.0 fps.

To address some earlier posts, I am not talking about dropping frames or anything like that. The video sources of the movies contain the actual frames that were shot by the film camera, and I am just saying to play those frames - all of them - sped up by 0.1% (during playback). Drawbacks could include “beating” with the refresh rate of your monitor, etc., but that’s another topic (and may be why this is not done typically).

Now, were those old cameras and projectors that accurate to begin with? Not sure.

And I know 0.1% is very, very small - I’m not arguing it’s really significant - this is more of a philosophical question. A “why not?”, if you will.

I’m not sure if you’re going to notice any difference between 23.97 and 24.

That’s a good question and could vary from person to person. But the question is really a “purist” one: should that step be done (if there are no drawbacks) to get as close to the original experience as possible? If changing the actual files would create problems (and I bet it very well might), should someone, in their home theater, change the playback speed on these movies by 1.001 times, or would that not be technically the right thing to do?

Over time, I’ve read a lot about film vs. video frame rates. The most confusing issue, perhaps, is the distinction between true 24p (24.0 fps) and NTSC video 24p (23.976 fps). As I understand, most films are shot in true 24p, and therefore I assume the original Star Wars movies were as well (especially given their age). I notice that, e.g., Harmy’s Despecialized versions indicate a rate of 23.976 when I play them in mplayer (Linux). This means the movie runs about 0.1% slower than originally filmed (and viewed in theaters). The audio is 48000 samples per second. And to match the 0.1% slower video, the audio is also slower and lower in pitch by that amount (I know it’s a small amount…).

I was playing around today with the -speed option in mplayer. If I do:

mplayer -speed 1.001 star_wars.mkv

I should be effectively seeing it in 24p, if all above is correct. I tried “-speed 1.3”, e.g., and it is noticeably higher pitch and faster, so it seems to work. Should this be the way we really should be viewing these old movies? If so, does it make sense to make video files with 24.0 fps and 48048 samples per second (or keeping 48000 but resampling, if that’s the only way)?

Just something to ponder…

towne32 said:

Looking at it a bit more closely (and remembering things better), the surrounding
shots are actually all from his 97 release, too. But the sequence in question just
happens to be unmodified from his 97 version. So, there shouldn’t really be a
difference in sharpness that needs to be corrected for those specific shots.

So are you saying that all of the Luke/Ben shots, from when Luke meets Ben to the shots inside his hut, are from 97SE? Or are you talking about a shorter part of that. The part that looks “different” to me from the rest is just from approximately the C3PO arm to the end of that scene right before the exterior of the hut (i.e. 00:31:45 to 00:32:13). The part from 00:30:11 (and before that as well) to 00:31:45 looks like 2.5 in terms of sharpness. It makes sense to my eye if 00:30:11 to 00:31:45 or so is from 2.5, whereas just after that is from 97SE. If otherwise (i.e. if you are saying all of 00:30:11 to 00:32:13 is from 97SE), then I am not sure what could be happening.

Note that where I really see the difference is in Ben’s beard, like at 00:32:02.

Which also partially answers your other question. The Blu-ray could be used,
but I thought Harmy’s take on it looked good.

So you mean 2.5 when you say Harmy’s take, right?

If there’s an issue with the 97 edit, it’s either specific to those few shots or
you’re only noticing it there. An enormous chunk of 2.7 uses the 97 edit,
more than just a shot here or there. Seems more likely to be something notable about 2.5.

I’ll see if I notice any other sections where there’s a shift from sharper to softer…

towne32 said:

This shot? http://i.imgur.com/HNimLm8.jpg

Yep, that’s the one: all shots in the scene from that C3PO arm to just before the exterior shot outside Ben’s hut. The scenes just prior (to the arm) look sharp, and the interior of Ben’s hut look sharp as well (well, except for the Luke lightsaber shot, which looks a bit blurry, but that one is blurry in 2.5 as well). BTW, I love the new lightsaber effect here - much more true to the original!

I think I used harmy’s 97SE for the rest of that scene, starting from that shot.

I’m curious: could the bluray source not be used for this scene (i.e. not try to fix the color in 2.5, but color-correct the raw bluray material? (forgive my ignorance if not).

The 2.5 color correction had been pushed so far that there was nothing left
(at least for someone with my skillset) to work with, especially in the final
shot of the scene.

Yes, and the specular highlight on C3PO’s arm in 2.7 looks much more natural. In 2.5 it’s very yellow.

Some of those shots, to me, look super sharpened in 2.5

Hmm, maybe they were. When you look at something sharpened next to something not as sharpened, it sticks out. It’s sort of a judgement call how much sharpening to use, but if the prior scenes were that sharp off the bluray, then keeping the same sharpness throughout probably makes sense. I did not go look at the bluray yet…

There’s nothing particularly bad looking about that arm shot, but you have to keep
in mind that it’s the same shot that pans over to the characters, which in turn
needs to match the other (same) shots of those characters as much as possible.

Yeah, all shots in that scene (from the arm to before Ben’s hut) look about the same softness to me. I agree they’d all have to be adjusted in the same way.

It’s a pain in the ass to color correct shots that are actually two shots, as such.
A huge example is Han getting up from Greedo’s table, walking up to the bar, and the
Death Star shot that it wipes to.

I bet. 😃

Anyway, I don’t think there should be anything wrong with the shot. But I
don’t know the specifics about the 97SE there. If something was done that
would reduce the picture quality or soften the image, or if the encoding was
somehow less quality (but I doubt it). I had the lossless files for that as well,
so re-encoding shouldn’t have been the issue.

Good you have lossless (which would make it possible, for example, to sharpen all 97SE shots you used to match surrounding ones, if that is the right move).

For other shots: If it’s a 35mm shot. Yeah, it’ll probably be softer.
It’s sharpened to a degree, but there’s a point where it starts to look bad.
It’s better than super-sharpened GOUT.

I agree GOUT is WAY over-sharpened. If, when from 35mm, you can sharpen to match both sharpness and grain of the rest of the footage, that might make sense, but I agree you don’t want to introduce too many artifacts if a match cannot be attained.

For the werewolf: Yeah, Harmy used the GOUT for that.
This was before we had super resolution techniques and,
more importantly, a good 35mm source.

Interesting, since I saw it on the werewolf first, and I dismissed it as an upsampled GOUT issue, but when I then noticed it on Leia’s hand, it made me start thinking it was some other effect. Might be interesting to check the bluray.

BTW, thank you for your hard work on 2.7 - there are some very nice enhancements there - it keeps getting better, and I can only hope my feedback is helpful in some way moving forward.

REPORT-SW-v2.7
00:31:46
This shot (which is only an example) in 2.7 is noticeably softer (less sharp) than in 2.5. Note that I have not gone through the whole 2.7 yet, so there may be additional shots (I think I remember seeing at least one or two others) that are also softer. Perhaps this is intentional, but nearby shots are sharper.

REPORT-SW-v2.7 (also, v2.5 - did not check others)
00:06:20
Red light illuminating Leia’s hand/arm looks pixelated/blocky.
Note I also see this in the red glow of the “warewolf” alien’s eyes at 00:44:09.
I am guessing it is triggered by or made more noticeable with saturated/pure red.