Slashdot Mirror
Motion Impossible: Tom Cruise Declares War on TV Frame Interpolation (theguardian.com)
An anonymous reader shares a report: At 9:46 last night, Tom tweeted an 87-second video in which he and his go-to director Christopher McQuarrie explained the concept of video interpolation and why it is the death of all good things. Video interpolation, they explained, is a digital video effect used to improve the quality of high-definition sport. "The unfortunate effect is that it makes most movies look like they were shot on high-speed video rather than film," said Cruise. "This is sometimes referred to as the 'soap-opera effect'." They explained that most HD televisions come with video interpolation switched on by default, they explained how to switch it off, and then they both nodded with total sincerity.
Now, it's worth noting that Tom Cruise is by no means the first film-maker to rail against motion smoothing. Back when he was still the Guardians of the Galaxy director, James Gunn tweeted that he, Edgar Wright, Rian Johnson and Matt Reeves were also peeved about the default nature of video interpolation, to which Reed Morano replied that she started a petition to fix the issue a number of years ago, to little avail.
Why did it fail? Possibly because none of these people are Tom Cruise. Because Tom Cruise has made a career of total commitment. Take him to a premiere and he'll spend hours on the red carpet, shaking every single hand until everyone's happy. Put him in a movie with helicopters in it and he'll teach himself to fly a helicopter to the level of a veteran stunt coordinator. Break his ankle on the side of a building, and he'll stagger out of frame on his ruined legs rather than blow a shot.
Now, it's worth noting that Tom Cruise is by no means the first film-maker to rail against motion smoothing. Back when he was still the Guardians of the Galaxy director, James Gunn tweeted that he, Edgar Wright, Rian Johnson and Matt Reeves were also peeved about the default nature of video interpolation, to which Reed Morano replied that she started a petition to fix the issue a number of years ago, to little avail.
Why did it fail? Possibly because none of these people are Tom Cruise. Because Tom Cruise has made a career of total commitment. Take him to a premiere and he'll spend hours on the red carpet, shaking every single hand until everyone's happy. Put him in a movie with helicopters in it and he'll teach himself to fly a helicopter to the level of a veteran stunt coordinator. Break his ankle on the side of a building, and he'll stagger out of frame on his ruined legs rather than blow a shot.
"Because Tom Cruise has made a career of total commitment." To Scientology.
"Eve of Destruction", it's not just for old hippies anymore...
This "motion smoothing" shit is the absolute worst. I would tolerate it at least slightly more if it ACTUALLY worked right. But it doesn't It'll work for 5 seconds, then turn off for 5, then on again for another 5. It creates a very jarring effect on the scenes. The software/hardware/whatever that is used to determine that one frame is related to another, so automatically splice in more frames CONSTANTLY fails.
Though, what I don't get, is that any TV I've seen in the past few years either doesn't have this "feature" enabled, or doesn't have it at all. I just purchased a brand new TV, a late 2018 model, and this feature doesn't exist. Other TVs I was looking at before this purchase didn't have it either. I think the feature died along with the 3D TV era. Which leads me to wonder why, now, of all times this complaint is showing up, since the feature is pretty much already dead?
Motion interpolation isn't great. But when they say "soap opera effect" that tells me that they aren't against motion interpolation, they are against high frame rates in general. This is analogous analogous to saying that 640x480 is the *best* resolution, and going higher makes things worse. I notice the article doesn't even mention the term frame rate. So this isn't a technical discussion, this is an aesthetic one.
Decades of watching movies has trained us to accept 24fps as "cinematic" motion, but in reality it just looks bad. 24fps is just barely on the cusp of fluid motion, and it gives some of us headaches. That's part of why video games consider 24fps unacceptable, as well as VR, and IMAX. Some people will say that it "takes getting used to" but it really takes getting "un-used" to the bad quality they shoot in today.
Motion interpolation should die. But the fact that people love it is signaling these directors that shooting in 24fps sucks and they need to move on.
Have to agree. For all his batshit crazy Scientology, from the little I have read, he doesn't push it on anyone.
Further, as the article alluded to, he does almost all of, it not all, his own stunts. Rappelling down buildings, skydiving, driving cars, leaping here and there, he's the one doing it. Not a stunt double.
Also, on those rare occasions I have seen him in an interview, he seems like a nice person. Maybe it's the Scientology, but he doesn't come off as stuck up or demeaning.
Give the man his due. He is accomplished. More than most likely anyone who posts here.
We will bankrupt ourselves in the vain search for absolute security. -- Dwight D. Eisenhower
How could motion interpolation ruin the lighting?
Watching *compressed* video is very difficult. I prefer to run it through a decompression tool first. One slight miscalculation on my Fourier Transforms and everything goes to hell. :-P
Perhaps broadcasts and streaming should contain in-band data to suggest to display devices optimal settings for the content. I don't want my day-time soaps to look like a Tom Cruise movie.
tone
In theory, you could do decent interpolation at the player since it has access to the MPEG motion vectors, and it can keep the audio and video in sync. Doing it on the video device is foolish because it has to essentially do the same work that the original compressor did, but in real-time, on a lower-quality version of the video (since it was compressed then decompressed).
Your LG OLED will have multiple display settings profiles.
You can set one up for general content and turn all that shit off. You can set another up for content where you do want it on.
Just hit the settings button on the remote, then change the picture mode to turn it on/off as appropriate.
On your LG OLED, pretty much the only feature you should enable is dynamic contrast, set to low. It gives non-Dolby Vision HDR content dynamic tone mapping, and makes shittyily-mastered HDR10 content (like many early HDR-enabled games) watchable.
Everyone uses SSL for their torrents and downloads these days. Comcast can't re-compress the files unless they can MitM the crypto.
Maybe 10-15 years ago you would have had a point about Comcast, back when they were considered "cable TV" but nowdays they're just another ISP.
I thought both Mission to Mars and The Red Planet were both pretty enjoyable movies to watch.
Avatar was also fun to watch.
I'm not surprised to hear this and used to agree, but there was an apparently-obscure movie a few years ago that I had to have called to my attention, and now I'll pass that favor onto you: watch Edge of Tomorrow. For two hours, anyone can stop hating Tom Cruise. And you can always go back after the movie is over.
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
Apparently the 24 FPS rate chosen for movies is an important psychological trick. It somehow tells the brain "this is fake", which makes the viewer fine with sets and costumes that would be cringeworthy when seen live. At 48 FPS, you lose that all-important filter, and everything is cringeworthy.
Socialism: a lie told by totalitarians and believed by fools.
it screws up the shadows. folded clothing has dark and light regions and when someting moves by it the interpolation just makes up weird new folds. It's a very hard to see artifact that is totally bizare when you do see it.
Some drink at the fountain of knowledge. Others just gargle.
Wow. Yeah, Forget the sex trafficking and slavery stuff the "church" arranges. Forget the threats of violence and intimidation. Forget that he is a principle financial backer for these scum bags -as long as he does his own stunts and "seems nice" (in fucking interviews? what a moron) it's all A-OK.
Compressed video...You get used to it, though. Your brain does the translating. I don't even see the code. All I see is blonde, brunette, redhead.
I think we should stop with that 24 FPS nonsense. It is not good, it is an artefact caused by technical limitations. Some people say it make films more film-like, that there is some artistic value in it, etc... I call bullshit, it is not a conscious artistic choice, it is a technical limitation.
If some directors chose to make some part 24 FPS like others chose to do black and white, then sure, that's art, but choosing that frame rate just because that's how cameras and projectors are setup isn't.
The only argument that makes sense IMHO is one of cost. More FPS is expensive: larger file sizes, more rendering time, a need for more sensitive camera sensors, etc.. Budget that can be better spent elsewhere. There is also value in having a standard, and 24 FPS isn't that bad a choice.
Frame interpolation is a work around that technical limitation, it is far from perfect but some people enjoy it. That's why TV manufacturers put it in here, film snobs be damned. Directors should learn from it instead of calling it "the end of all good things": many people want smooth motion.
That being said, I am totally fine with 24 FPS, and I don't use interpolation, but just be honest and say it is preliminary a cost saving measure, with maybe a hint of nostalgia.
I recently got a new LG 65" OLED.
First thing after turning it on, was to go into the menu system and turn OFF all the stupid auto-correct stuff, including the motion smoothing stuff.
I then turned off factory "torch mode"...and began adjusting the colors to be a bit more realistic and cinematic.
ON thing I was a little concerned about, was it was a little dimmer than my plasma it replaced. I had read about this.
But then, I found an "ECONO" mode and turned that off and WHAM...the screen got way brighter than I needed and I had to turn it down.
So, if someone is telling you that OLED can't be as bright as the QLED (Closest competitor)....they may have not discovered turning off the econo-mode.
But yes, it is sad that you pay this much $$$$ for a good televisions and by default, it look like shit out of the box and you have to manually fix things.
Even sadder...most people do not do this and they look at a very expensive crappy picture.
Light travels faster than sound. This is why some people appear bright until you hear them speak.........
HERE is a really great explanation of frame rates starting with movies and spreading and co-existing with TV.
It actually had to do with when silent films went with sound....and budgets and $$.
Many changes over the years, but once a standard sticks, well, it is hard to change.
HERE is another video by the same guy, of "in defense of 24fps"....as to why it likely will be around to stay for a long time.
Light travels faster than sound. This is why some people appear bright until you hear them speak.........
The last 10-15 years he cranks out endless derivative action films and sequels.
That's fine, that's your opinion. I think Oblivion, Edge of Tomorrow, and Ghost Protocol are absolutely fantastic films. Now The Mummy on the other hand...
Meh, if Tom Cruise thinks TV Frame Interpolation is a bad thing- then I'm all for it. Put it in everything I say! Even cheeseburgers.
I believe that is called "cutting off your nose to spite your face."
This is going to sound weird, but I'd like to see the concept of "frames" disappear entirely. Although it'd be a pretty radical change. I'd like pixels to follow piecewise spline curves with their start/stop times being at arbitrary floating point values.
The data would effectively be captured thusly for each pixel**: the first part of a new piecewise step would be used to determine the curve shape, and then no action would be taken until the deviation from this curve exceeds a breakout threshold - the point where no adjustment to the curve shape can accurately described the data gathered thusfar. This curve accumulation / breakout be done in hardware, atop the CCD layer**. Recorded pixels** would increase or decrease the breakout threshold of their neighbors, in order to encourage whole blocks of pixels to transition between splines at the same time (for compression reasons - you wouldn't want to have to record a header spelling out the coordinates** and start time for every pixel** individually). A step between splines might be so fast that you have to watch at 1/1000th speed slow motion to even see it - or it might last for seconds. It all depends on the scene.
Note the asterisks (**) in the above paragraph. Because rather than referencing pixels by the x,y coordinates of a CCD pixel, one would ideally have a layer of separation that maps CCD pixels to fixed polar coordinate positions centred around the camera's focal point ("virtual pixels").This would let you shift the CCD-polar coordinate mapping based on the camera's accelerometer data, so if the camera is rotated, the virtual pixels still correspond with the same real-world object (e.g. slewing the camera doesn't invalidate all your splines). Virtual pixels in polar coordinates would also support full 360 recording and playback.
The video file format would be grouped into blocks (each sharing a single start time) containing clusters of pixels (each containing metadata describing what run of pixels you're updating), followed by each pixel's spline data (in a compressed format that makes use of data correlation between adjacent pixels). The more the compression is desired, the more it fudges the start times to group together larger blocks. A player just reads through the blocks, waits for said floating point start time to occur, then updates the splines for all pixels described therein. The screen displays whatever splines are in its memory, at a hardware level.
All blocks could also be tagged with a camera ID, and camera metadata (containing said camera's coordinates and orientation relative to some fixed coordinate system) could be periodically provided. This would allow different cameras to record the same scene simultaneously and be played back simultaneously. This would allow, for example, stereoscopic 3d, or for the data to be used in actual 3d scene reconstruction. I'd also love for information about the frequency bands recorded by each cxamera to be stored in metadata (with any number of frequency bands allowed, rather than just a generic "RGB"), so multispectral imagery could be recorded and reconstructed.
To me, something like that would be the ultimate recording / playback system.
Seen on a Japanese food processor: "Not to be used for the other use."
https://www.blurbusters.com/fa...
There is more to the motion blur problem than the slow transition times of LCD pixels. Newer monitors have much faster transition times and the problem is still there.
What I think is happening is that the CRT is producing a kind of impulse sampling of the moving image whereas the LCD is producing zero-order hold (square-step, see https://en.wikipedia.org/wiki/...) output. The human visual centers appear to perceive the "strobed" image of the CRT as smooth motion, the "change-and-hold" image of the LCD as blurred, even at high frame rates and with rapid pixel response.
The reason I say "kind of impulse sampling" is that the CRT does not flash a sequence of static images the way a film movie projector does. Rather, the CRT conducts a continuous raster scan, with a short blanking during the retrace. Each line of the image gets strobed at the time the scan reaches it, but each line is strobed at a different time instead of the whole image all at once as with a film projector.
I believe it is that scanning that accounts for the "soap opera effect" of video content recorded on video tape instead of on film. This is already a long while ago that a local TV station had a show-and-tell of this new thing called HDTV at our Engineering campus. The Engineering profs were oohing and ahh-ing about what they thought were amazing images, but I was pointing out the image artifacts (easier to spot in HD!) to the broadcast engineer from the TV station, and finding a receptive audience, he went on at length to explain the difference between Homicide, Life on the Streets, shot on video tape and having the soap opera look, compared with Law and Order, which he explained was shot on 35 mm film and then scan-converted for TV broadcast.
So, even if the CRT scanned mode of projection differs from the flashed-image mode of film projection, apparently recording the image on film, which records a sequence of still pictures, has a better look than video tape, even when film is played back on a CRT.
The other problem is that most people viewing video think that HD on a widescreen LCD looks fantastic and don't know what us motion-blur worriers are complaining about. This population includes engineers developing TVs and computer monitors. The only people complaining, it seems are hard-core gamers along with people who have seen the Kay 5500 Sonograph http://jproc.ca/rrp/sonagraph_..., a scientific instrument used in speech science that used a DSP to drive a CRT (at VGA resolution!) that produced a truly remarkable visual effect of a "voice print" rolling past the screen with zero motion blur -- the later software spectrum analyzers producing un-synched scrolls to LCD monitors of much higher frame rate look terrible by comparison.
With respect to the awful motion blur of LCDs, which other posters here is telling me in not cured by video interpolation, there is an element of what Robert X Cringely described in Accidental Empires, when (back in the day), a techie gushed about the desktop publishing revolutions, showing off the font quality of LaTeX printed at 300 DPI on a LaserJet II, which Cringely looked at in dismay in comparison to what the publishing industry got from photo typesetting.
DPI and frame rate are important, but if the community is at all serious about further advances in video, especially VR, engineers are going to have to take the physiology of human vision and the motion blur problem into account.
"The unfortunate effect is that it makes most movies look like they were shot on high-speed video rather than film,"
That is easy to explain it's because they were shot on high-speed video rather than film.
So why not just record the movies at 120 frames per second? Then there's nothing to interpolate
So how often do you sample the CCD? That sampling rate is your frame rate. And keep in mind that today's "frame" is an accumulation of all the light hitting the sensor since the last frame. To sample more frequently, you get less light and a noisier image. Yes, you can do something like a "rolling shutter" but there are limits before it starts messing with motion.
The real issue is that the interpolated frames wake up your body thetans, but Tommy doesn't want to get into that, because only people who have fully paid up for OT III are allowed to have this knowledge.
By your logic, every Catholic should be burned at the stake because they're a backer of the world's largest pedophile ring.
We will bankrupt ourselves in the vain search for absolute security. -- Dwight D. Eisenhower
Why it was chosen in the first place is different from why it is good. The reason film has stayed at 24 FPS for a century isn't technical.
Socialism: a lie told by totalitarians and believed by fools.
Well no consumer watches uncompressed difgital video, i suspect a few hollywood okorists gradesndrharfor watch uncompressed, and maybe the studio boss for finsl sprowal befor final oput to didital cinama / film (for cinemas nor gone digitak yet and all the other distribution formats consumed by evrybody else
I think you may have a flipped bit in your decompression.
Haven't seen it myself, but some people don't like the look of high frame rates for movies, apparently gives it an unnatural look.
I remember seeing the last Hobbit movie in the theater projected at a full 48fps. I remember thinking to myself the video looks odd, like a PBS re-enactment video shot in really hi-res. Yes it was smooth and clear, but it just seemed very unnerving and I was sucked out of the plot a few times because of this. I have since seen this on my tv at home at 24fps, much less distracting.
I have also been to people's house where they have the 120fps smoothing turned on and it too looks odd. The motion seems a bit jerky like its speeding up and slowing down. When I see someone walking with this turned on it looks like someone repeating the silly walk sketch form MP.
So why not just record the movies at 120 frames per second? Then there's nothing to interpolate
It's too fast at 120 FPS. Just drives up costs for no real benefit. You cannot see much more than about 50 FPS at reasonable distances.
Movies to film where traditionally shot at 24 FPS, even IMAX film is shot at that rate. Standard definition TV was 30 FPS interlaced. The biggest issue here is that FILM has way better resolution than Video, but runs as 24 FPS instead of 30 FPS. Translating from 24 to 30 is not an easy bit of math, so there are a number of schemes to deal with it.. Usually you just duplicate film frames every so often to bring 24 FPS up to 30, some just run the film at 30 FPS but it looks weird (think Charlie Chaplin walking in black and white, it's too fast and looks strange because old silent films where shot at even lower FPS.)
I actually find that old "film" based movies don't display well in HD or 4K, even if shot in 70mm. Most of these feature films did not have the production quality to support higher resolutions and I find myself being distracted by the in appropriate set detail or costumes and special effect artifacts that wouldn't be visible on a DVD. I remember the first time I saw the first "Pirates" move in HD from Blu-ray, it was horrible.
Frame rates of 120 FPS are about 3 times what you actually need as a frame rate. You cannot see much more than 40, though eye strain may be an issue. The way to avoid that, is to use 40 FPS frame rate, but scan it at 120 FPS (i.e. show the same frame 3 times). They actually did this with film projectors, where they'd flash the same frame multiple times.
Recording at 120 FPS may sound neat, but the problem is it simply isn't worth going above 40 or 50 regardless of the material. Higher frame rates simply drive streaming bandwidth up, storage sizes up and production costs up, but add no perceived value to the end customer. Resolution though, IS worth it, if not now in the future.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
But then you encounter the real thing* and fail to "perform" because your mind expects compression/interpolation artifacts during the act.
Took me years to get over ASCII "content". She wouldn't wear my special symbol stickers.
* Then again, this is Slashdot probabilities we are dealing with.
Table-ized A.I.
Why it was chosen in the first place is different from why it is good. The reason film has stayed at 24 FPS for a century isn't technical.
No, perhaps it's not, but it's financial which is a bigger driver than technology in the movie business. If you have millions worth of equipment that uses film, you use it until it doesn't do the job anymore.
16mm film is good enough for old TV, 35 mm film is better than 1080p/i, 70mm is even better than 4K. It runs at 24 FPS though a host of existing processing and editing equipment. Folks know how to use it. It's cheap...
However, that's not to say that new production companies are not being created and equipped with HD video equipment or that the editing chain isn't quickly being converted to digital, it is. The issue is that capture of film is so well understood and high resolution video camera systems and lenses are not as advanced or as inexpensive, yet... I've seen that some of the new media companies (say Netflix) that are producing their own content are dumping film production for the quick turn that digital gives them now that distribution on film isn't as common as it was. But these companies are not producing material to project on a 100' wide screens to start with, so they don't really care about film at all.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
The phosphor in a CRT is not a glowing filament with a longish thermal time constant. A closer analogy would be a fluorescent bulb on a magnetic ballast, and those things flicker like crazy. A dude I know, who worked for a state agency promoting energy saving, would visit schools and give out these cardboard wheels you could spin to show how a compact fluorescent on an electronic ballast operating at a much higher frequency didn't do that.
The support for my hypothesis is comparing a retrace-synched scrolled image on what back in the day was a decent Sony Trinitron CRT 1) at 60 Hz refresh, 2) 120 Hz refresh and 3) 120 Hz but doubling each image to simulate the persistence of an LCD screen at 60 Hz. 120 Hz scrolls blur free, as does 60 Hz but with noticeable flicker on the Sony monitor, but the "double strobe" at 120 Hz simulating a 60 Hz image refresh is unmistakably blurred.
The Blurbusters agree with me in that they strobe an LED backlight of an LCD to suppress motion blur by simulating a CRT. Much of the discussion is how some really expensive displays intended for vision research do just that, and how this strobing the backlight can be done more cheaply by some hardware hacks. They are not talking about higher screen refresh rates, not about image interpolation, this is just plain getting rid of the sample-and-hold square-step effect.
The problem is akin to the fractal coastline problem: How long is the coastline of a country? The answer between "theoretically" and "actually" is different. Theoretically, it's like a Koch snowflake - the perimeter is infinite, because each time you zoom in, you get new curves superimposed upon the curves. In practice, you hit real-world limits due to a change in physical processes at small scales.
Ideally, there's no practical limit to how fast you could sample frames. Indeed, an ideal implementation would be direct spline accumulation, with readout triggered only when the threshold is exceeded (that is to say, the act of photoaccumulation itself builds the spline, with inhibition occurring between the different spline parameters corresponding to their exponents). In such a case, you'd be limited by nothing more than the minimum number of photons required to build a valid spline to the point of it exceeding its threshold.
With a standard CCD like we use today, however, where you have to read out rows at a time with no individual per-pixel logic, you have to pick a readout speed. The highest acceptable readout frequency is based on how much light you're receiving in your scene. That said, even on consumer-grade cameras, in outdoor daytime situations readout speeds measured in the thousandths of a second can be quite acceptable. True, these are not "infinitely short" periods of time, but they sure are pretty dang short. To put it another way: the "worst case" is what we do now. The "best case" is vastly better.
To point out a specific advantage: Super slow motion is a popular feature on cameras these days, but you generally only get short bursts at limited resolution. You're accumulating entire frames of raw data - the vast majority of which could be described by only a few splines throughout an entire burst, with only the most active areas requiring a meaningful number of splines. But rather than collecting a small number of spline parameters, it's reading out huge amounts of pixel data every frame, and it must store all of it. This limits A) maximum slow motion framerates, B) slow motion burst lengths, and C) slow motion resolutions. Spline accumulation would face none of these limits because the data stream you're accumulating would be vastly smaller (assuming its done directly in the sensor during CCD readout - there's no advantage if you have to buffer all the raw data and then process it).
Seen on a Japanese food processor: "Not to be used for the other use."