Hacker News Reader

Comment by bovine3dom | Hacker News Reader

This whole thread and article is like it's from an alternative universe to me. All of my tech-savvy friends go to great lengths to turn motion smoothing on everywhere ("smooth video project" is the best I know for PC) they can so that panning shots don't look awful. They wouldn't play videogames at 24Hz, so why should they watch a film at 24Hz?

It does introduce artefacts. The solution to all of this is for filmmakers to move on from the 1930s and film at 120 fps or higher. If people only want to watch a quarter of the frames, it can be downsampled easily.

The argument reminds me of musicians who didn't want their albums chopped up and sold as individual tracks, to be consumed however listeners desired.

View on HN

Panning shots probably look awful if you're watching 24 fps content on a device running at 60Hz (many consumer devices, computers, and Blu Ray players do this by default). Since 24 frames can't be evenly spread out over 60 frames, you'll see motion judder, and it does look terrible.

However, 24 fps running at a true 24Hz is perfectly fine, and you will not see flicker - at least not if produced by competent filmmakers. That's because the camera's shutter angle is normally set such that motion blur makes the motion look smooth. Real cinematographers also know exactly how fast they can pan with a given lens without causing any noticeable strobing motion.

Comparisons to video games completely miss the point. Most video games are unable to properly simulate motion blur (some try, but it doesn't usually work well), so you have to have high frame rates for things to look smooth. Since you need to react quickly in video games, high frame rates (combined with the lack of motion blur) are also helpful in preserving crisp edges on moving objects (so it's a practical advantage). And finally, video games generally try to simulate reality for the player, so players are more concerned that the technology makes the experience believable rather than intentionally suspending disbelief (as in film or theater) to passively unpack the narrative nuance of artwork unfolding on a screen.

For those same reasons, sports also do well with higher frame rates, but when fiction (which is obviously disconnected with the present reality) tries to use high frame rates, it falls into "uncanny valley" territory - much like the computer-generated humans of movies like "The Polar Express". As others have noted, a few directors have really tried to push HFR film to the public, but it has never been received well (whether or not HFR is recognized as being responsible for the uneasiness felt by audiences watching HFR content).

As for HFR content being "easily" downsampled - it really isn't, at least not with respect to motion blur (which is an essential point that many people miss in these discussions).

24fps pans don't flicker, but they do easily lose the sensation of motion and look like a series of static pictures, even with perfect zero-judder frame timing. This can feel a lot like flicker, and even if you can perceive it as motion it's still very bad looking motion.

Closed down shutter angles are only needed to mitigate the blur of low frame rates. If you're shooting at 120fps you can use a fully open shutter (which helps with the lighting requirements too), which gives you a choice of 5 different simulated shutter angles when you downsample to 24fps by blending frames. Blending 2 consecutive frames then deleting 3 gives you a look very much like modern 24fps film.

I'm not talking about 3:2 pulldown; I'm exactly the sort of sad sod who has spent the time writing scripts to match my display's refresh rate with the media that was running. I strongly dislike watching panning shots at 24Hz; I do not think it looks perfectly fine.

You're right about motion blur making 24fps look less awful, though. I'm sure you know that filmmakers often reduce it in fight scenes to deliberately make viewers a bit uncomfortable. Thanks for pointing out that downsampling wouldn't work without some extra effort (e.g. just having the same amount of blur but with the extra frames such that when it was downsampled it looked normal), as that isn't something I had considered.

I find the suggestion that videogames don't need to suspect disbelief, but films do, very strange.

> I find the suggestion that videogames don't need to suspect disbelief, but films do, very strange.

True, the distinction may seem strange to some, and it's not always a clear distinction. Here's how I'd describe it in more detail (just my opinion of course):

In film, we see a person depicted on the screen with certain characteristics that convey emotion (just as we do with a group of lines comprising a stick figure with similar emotion), but we are forced to activate parts of our brain that span the gap between what we see (which obviously isn't reality) and what that character might feel in his/her universe.

To some degree that same experience happens in video games, but I would argue that instead of seeing the emotions and trying to understand them, we (as players) are the person feeling the emotion. In one, we are witnessing events happen to someone else (and exercising empathy), whereas in the other, we are experiencing the events first-hand (even in a 3rd person view, we're playing as if we are the character in view).

A first-hand experience feels most believable when it's actually realistic (hence the need for high frame rates), but a simple stick figure can feel "believable" if we're engaged and sharing the emotion we believe it to be feeling.

As a side note, maybe that's why I personally prefer classic (8-bit era) video games over modern 3D video games :)

>As for HFR content being "easily" downsampled - it really isn't, at least not with respect to motion blur

Er, why not? Motion blur is simply time integration. Why can't you simply average adjacent frames during your framerate decimation?

In fact, since "real" motion blur from shuttering uses a likely-suboptimal square windowing function (shutter is binary open/closed), I'd hazard it possible that motion blur computed from a higher source framerate could potentially be more perceptually pleasing (by using a smoother windowing function). You have more information to work with, after all.

> Why can't you simply average adjacent frames during your framerate decimation?

In theory you could, but in practice, it doesn't work very well and you tend to get some really nasty artifacts depending on the motion of the scene. There's a lot more to it than simply averaging adjacent frames since those source frames likely aren't capturing perfect 360° shutter angle time slices that seamlessly align (temporally) back-to-back.

Ang Lee already tried the 120fps thing, it was a colossal flop. No one likes watching movies in anything other than 24fps dude.

Filmmakers filming in 120fps would destroy the film industry overnight because NO ONE would watch what they produced.

> No one likes watching movies in anything other than 24fps dude.

I know of at least 6 people who prefer higher framerates, myself included. I don't know anyone who has expressed an opinion the other way. As I said, this whole thread is like some kind of alternative universe to me.

Motion smoothing apparently hasn't stopped people buying TVs.

Many people hate motion smoothing though; my mom was complaining that their new TV made everything look like a soap opera until I disabled it.

Like many other things, this is all down to personal preferences; similarly I have friends who can't stand to game at anything <120FPS and one who gets motion sick at 60FPS, while I would rather turn up the graphical detail and survive with 60. Neither opinion is more correct, and you have had several people upthread saying they prefer 24 for films (myself included).

You've missed the part where 120 is divisible by 24. You can losslessly go from 120 to the other popular frame rates, including 24.

Edit: nope, recreating from higher fps the motion blur that's common in 24 would be way more expensive.

> You can losslessly go from 120 to the other popular frame rates, including 24.

Not true. Something filmed natively at 24 fps and a 180° shutter angle is vastly different in terms of perceived motion than 120 fps converted to 24 fps.

Are 1/48s exposures really that common in cinematography?

Yes.

"By far the most common setting for cinema has been a shutter angle near 180°, which equates to a shutter speed near 1/48 of a second at 24 fps." [1]

"A 180° shutter angle is considered normal." [2]

[1] https://www.red.com/red-101/shutter-angle-tutorial

[2] https://en.wikipedia.org/wiki/Rotary_disc_shutter

Oh.

Right, replicating that exactly requires 240fps with fast shutter and discarding half of the frames. Which is a way more expensive proposition.

If you shoot 120fps with 360° shutter angle you can downsample to 24fps with 144° shutter angle, which is close enough to the traditional 180°. And in any case cinematographers often vary the shutter angle, so viewers are already used to changes. Narrow shutter angle is common in modern action movies.

Yes, cinematographers do vary the shutter angle for artistic purposes, but that just confirms the fact that the amount of motion blur present in a film is a deliberate artistic decision rather than a side effect of using an "antiquated" frame rate.

As for downsampling, I believe your math is off. 120 fps at 360° means each frame is exposed at 1/120s. Conforming that to 24 fps gives you the same exposure duration per frame (you're just throwing out 4/5 frames), so you're effectively at a 72° shutter angle - less than half of what's considered "normal".

I only said that 144° effective shutter angle is possible, not that you can achieve it by deleting 4/5 frames (which does give you effective 72° shutter angle). You actually need to delete 3/5 and blend 2/5 (and the 360° shutter angle of the 120fps source needs to be accurate enough that you don't see the join).

Reply on news.ycombinator.com