Wednesday, December 5, 2012

48 Frames Per Second NOT What Makes Hobbit Look Like Soap Opera Video

It's all you read this week:  Peter Jackson's decision to use "48 frames per second" rather than the cinema industry standard 24 is what has single-highhandedly made The Hobbit look like a daytime soap opera video, where you can tell everything is just a movie set, and not a "real" world in which hobbits roam.

Nothing could be further from the truth.  Faster frame rates have been around for decades. Yes, decades and no one has ever said they made film (I'm talking real film) look like soap opera.

Proof?  I just finished producing and watching a 60p HD video and it looks exactly the same as the 24 and 30 fps versions except that it lacks a bit of the jitter.  It does not look like a video game, nor do moving subjects "float" above background as if they were cardboard cut outs.

It it's not high frame rates per se, what is the culprit?  Most likely it's motion smoothing of some sort.

Walk into any big box store and you will see TV's lined up with special "demo" settings turned on to help them fly off the shelves.  The brightness is way up, the colors over saturated, and a feature called motion "smoothing" turned on.  The net result is a Hollywood movies looks like a video game.

It is the motion smoothing algorithms that are to blame, not a high native frame rate.  Any motion smoothing, whether cheap consumer grade or "pro" grade, works by a computer trying to guess what additional frames in between the actual one in a video or movie would look like.  If you shoot in 30 frames with a ball moving across the shot it will attempt to fill in between frames.  It will literally take information from two frames and interpolate approximately what the ball would look like if there were an additional frame between them.  But it is far from perfect!  Often a black outline appears.  Or the motion is too oily smooth for the surrounding video.  It makes the moving subject "float" unnaturally compared to the background.

Rendering additional frames can also give a video look of a video game, or make a Hollywood film look like it was shot on a Handycam.

Peter Jackson has been mum about any digital smoothing. But the effect is clear.  One recent CNET editor put it:

I noticed that artificial look at one point early in the film, when Bilbo gets astride his first pony, and the background moves behind him in a scroll, as if it were printed on a piece of paper unrolling to simulate his riding forward.
This separation of forground from background is a classic artifact of digital smoothing programs such as Twixtor.  High frame rate, however, does not cause that type of artifact.

You can see this even at 30 frames per second on commercials on your TV when it was applied with crappy "motion smoothing" effects during editing.  The edge sharpening that makes things look "too real" survives reduction to 30fps for broadcast. 

Take for example those wretched Sandals commercials. Once you get over gagging on the overuse of Time of My Life,  you see couples enjoying the beach then suddenly for a second the speed is increased.  In the past, the look would have stayed the same, just faster.  But not with crappy "motion smoothing," or frame interpolation. Suddenly it looks like pre-digital HandyCam video dropped into a high end commercial.  But again, it's not the frame rate per se.  It's the digital motion smoothing processing.

What do digital smoothing filters on TV's do?  Here's a description of how LCD HDTV does it:
Most LCD-TV vendors have introduced technologies that generate new frames of content based on interpolation of the source material. The programs use mathematical algorithms to analyze adjacent frames (or a group of adjacent frames) and figure out where moving objects might logically be in the fraction of a second that the source video doesn't contain.
So a consumer grade chip in the TV is finding things that move in the scene, isolating them, creating new fake "in-between" versions and the result is they very smoothly float like a paper moon over a cardboard sea.

The result is that a cow walking through the scene is outlined,  cloned and "tweened" between the first real frame and the next real frame.  Yes, now the cow smoothly floats across the screen with it's body not moving naturally on it's legs.

Another proof that frame rate alone does not remove the film look:  take 24 fps and reduce it to 20 or 12 fps and ask:  does that look more film like?  No.  It simply looks a bit more choppy.

Picture an old flip book animation in your hands.  Flipping the pages at 48 per second won't suddenly make it look real or a foreground object "pop" off the background.

Yet all we read on Jackson's Facebook page is that it was "48fps" that accounts for everything.  And the press piles on with pitchforks and torches:  "Death to 48!" 

More journalistic silliness abounds with some suggesting it's because he shot on, digital rather than  film.  But that just doesn't wash.  There are plenty of all digital films that have a wonderful filmlike look. Not one has noticed. But we read things like this from Jason Gorbor on Twitchfilms (emphasis mine):
From the music to the costumes to the iconic New Zealand vistas, it's easy for any fan of the other films to immediately feel that they're returning back to middle earth, except for one major technical change - eschewing the celluloid used as the main capture format for the previous trilogy, Jackson has instead shot The Hobbit on digital video, in 3D, and using 48 frames per second ("high frame rate", or HFR, as opposed to the normal 24fps), and with a 270° shutter angle.
Then there's the argument that 24fps has a lot of "smearing" and "motion blur" and that's what makes film look like film.  That doesn't even make sense.  Shutter speed versus object's motion alone determines blur whether you're shooting one still or 1000 frames per second high speed.

And that brings up another point:  what about all those ultra high speed shots we've seen of bullets piercing an apple, or of explosions on Mythbusters?  Even at 1000 or more frames per second, whether you play them back at 24, 30, 60, or 1000 fps they don't change their look.   In fact, they look rather film like at any speed! 

And forget "jitter."  Jitter/judder doesn't mask a movie set's fakeness any more than scratches on the celluloid, or gate weave do.  They exist in a whole different realm of artifacts that have nothing to do with the presence/reality factor of film.

Jackson uses a very unfortunate analogy to explain the "new" effect of his HFR 3D on the audience:
HFR 3D is “different” — it won’t feel like the movies you’re used to seeing, in much the same way as the first CDs didn’t sound like vinyl records. 
Oh wow!  Does anyone remember listening to the "first CD's" on the first CD players?  Ouch!  They sounded awful.  It wasn't about getting used to it.  They still sound wretched if you were to drag out that old Phillips 1984 CD player.   It wasn't "different"--it was defective! 

Motion blur.  Here's another demon Jackson criticizes.  He implied that somehow 24fps always has tons of motion blur and that always obscures details.  But this is wrong for several obvious reasons.  First, many 24fps scenes have little or no motion!  And "detail" does not suddenly increase in a shot that goes from pan to static where the audience says, "Wow, suddenly it looks so real."  Secondly it all depends on your shutter speed or angle in cinema terms whether or how much motion blur you get.  Sure, if a cinematographer insists on shooting everything at 360 degrees (i..e., the full 1/24 of a second a frame is in the gate to be exposed) or, in Jackson's case, 270 degrees (3/4 of the max exposure time possible for a frame), then, yes, I supposed 48fps would make less blur.  But for action scenes most filmmakers choose a much faster, smaller shutter angle.  And even then, motion blur from the relatively slow shutter speeds of a cinema camera actually helps to smooth motion naturally, making it look less staccato.  And 270 degrees with 48fps is still only 1/64 exposure time, slow in the world we are used to with DSLR's.


  1. I don't know how I stumbled upon this, but as a DoP and an Editor I can say unequivocally and without doubt that the look is not only directly attributable to the frame rate, but the 'video look' can easily be reproduced on all the cinema cameras you mentioned. The standard shutter angle (which is 180º, not 360º) produces motion blur relative to the frame rate. Shooting 24fps with an angle of 180º gives you exposures of 1/48th of a second. Shooting 48fps with a 180º shutter gives you exposure of 1/96th of a second. Motion blur is significantly decreased between frames and as a result detail increases (which is why the sets and make up looked worse on the big screen, the motion blur of 24fps is a significant factor in decreasing sharpness without a 'perceived' loss of sharpness).

    Your comparison with 30fps and 60fps video is also flawed, because 24fps is considered a 'low' frame rate, whereas 30fps is a 'medium' frame rate, and thus has a different 'look' to 24 (and PAL 25) fps. The example video shown doesn't even say what shutter angle it was shot at - thus negating the 'results' as relevant to your argument.

    The fact is, the soap opera look comes from shooting interlaced video, which 'technically' is 50/60fps. In camera, interlaced and HFR progressive modes look very similar when monitored because they're essentially the same thing, albeit interlacing has half the temporal data per frame and is a pain in the butt in post.

    I'll cover another part of your argument - stop motion animation as it is a bit of a hobby of mine. Stop Motion has issues to do with frame rate and motion blur just the same. Ever notice that cinematic stop motion is rather juddery? A lot of motion blur is also added in post to try and make the appearance more comparable to film. Most stop motion is animated at 16fps (even Aardman use 16fps for their features I'm pretty sure), and boy can you *tell* in a cinema. This ties in nicely with your idea of a flip book of photographs and also with shutter angle again - 24 photographs at 1/96 will look juddery and rubbish, but very sharp like photographs, and adding another 24 frames to it will then give a very 'soap opera' type look. But if the original 24 frames have a 1/48th shutter and you add 24 more with the same angle, the look will still be 'film like' because of the blur.

    *a 180º shutter allows time for a full exposure of a frame of 35mm film, and then enough time for the next frame to enter the gate. Not so relevant for digital today where 360º shutter is possible, but in the film days 360º would imply that the shutter was permanently open and then you'd get some really nasty looking film.

  2. Justin, thanks for your input. I notice that examples of a Canon DSLR with a shutter speed of up to 600 do not create the video look however.

    My point with the flip book and actual high frame rate film (real film, not digital) is that no one ever said in those days it looked like video, made the sets look fake, or made moving objects pop out from the background. That is something that has only, to my knowledge, ever been reported since the digital film era began a few years back.

    Yes, slower shutter/angle will introduce blur and create better looking stop motion, and hide detail, but if lack of that was what created the extreme "pop out from the background look" then why do all those classic super slow motion shots of a bullet going through an apple not look like video or the extreme pop out look you see on the current crop of flat screens in Sams that make 24p movies look like an elementary school play, with a flat painted background even when the original background was a real scene and shot in 24p? Just saw it again last week wit OZ playing on a flat screen. Franco looked like he was standing in front of a poorly lit theatrical flat....and yet it was a real outdoor scene. That's crummy digital processing or "smoothing" they use to sell sets to consumers.

    Now I'm going to have to buy a few rolls of film again and make a flip book and test it :>