Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
Movies Entertainment Technology

How 'The Jungle Book' Made Its Animals Look So Real With Groundbreaking VFX (inverse.com) 152

Posted by msmash from the pushing-VFX-to-the-limits dept.
An anonymous reader shares an article on Inverse that looks into how The Jungle Book movie was made. Following are some of the interesting tidbits from the story: Directed by Jon Favreau, this version of The Jungle Book, which borrows from both Disney's 1967 cartoon and the original Rudyard Kipling novel, sets a new standard for life-like CGI animals. Shot entirely on a soundstage in downtown Los Angeles, it is sort of a hybrid of Avatar and Who Framed Roger Rabbit, with one human performer surrounded by animated creatures -- the difference being that every effort was made to trick the audience into believing the animals were real. [...] For the most complicated scenes, the computational power required was astounding. "It would take 30-40 hours per frame, and since it's stereo [or 3D], it requires two frames to produce one frame of the movie -- at 2K, not even 4K," Oscar-winning visual effects director Rob Legato said. "So you can tell how much the computer has to figure out, exactly what it's doing, how it's bouncing, how much of the light is absorbed, because when it hits an object, some gets absorbed and some gets reflected." The math there is mind-boggling; it takes a full 24 frames to make up a single second of the movie, and most shots are between five and ten seconds. That required "literally thousands of computers," Legato said, and eventually, some creative solutions. "I think they started using the Google cloud, which has tens of thousands of computers, and sometimes it would take two or three days to render a shot, he said, exasperated at the mere thought of the process. As powerful as the computers were, they ultimately were just taking cues from the human innovators who spent years on the film. "In all this," Legato said, "there's no real computer that replaces the skill of the operator, of the person who is pushing the buttons."
This discussion has been archived. No new comments can be posted.

How 'The Jungle Book' Made Its Animals Look So Real With Groundbreaking VFX More

How 'The Jungle Book' Made Its Animals Look So Real With Groundbreaking VFX

Comments Filter:

  • How's the story?

    • Re: (Score:2)

      by haruchai ( 17472 )

      Ask Kipling.

      • Re:That's nice. (Score:5, Funny)

        by jratcliffe ( 208809 ) on Monday April 18, 2016 @12:27PM (#51932957)

        Do you like Kipling?

        I don't know, I've never kippled.

        • Re: (Score:2)

          by KGIII ( 973947 )

          So, I herd u liek mud kipling?

          Err... Let me just apologize in advance for that. I'm sorry - but not sorry enough to not do it.

          On a more serious note, I bet the rendering farm for this was AWESOME!!! Holy crap... I've seen some previews for the movie and the first thing I thought of was the rendering farm. Wait, no, the first thing I thought of was the original. The second thing I thought of was that I was singing "The Bear Necessities" aloud. The third thing I thought of was probably the rendering farm. I a

    • Re: (Score:2, Interesting)

      by markus ( 2264 )

      I feel, this is probably one movie I don't need to watch. As you said, the story is likely to take a back seat to the visual effects.

      And maybe, it's just me but seeing the previews, it felt extremely jarring. The visual effects are so much on the wrong side of the uncanny valley. It is glaringly obvious that all the animals and a lot of the background is CGI rather than real objects. And to me it feels very disturbing and distracting. Of course, if there is no story to distract from, then maybe that doesn't

      • Re: (Score:1)

        by Anonymous Coward

        [It's glaringly obvious that all the animals aren't real.]

        You mean aside from when they talk?

        • Re: (Score:1)

          by Anonymous Coward

          You mean aside from when they talk?

          I know. I felt the same thing while watching the version from 1967. The animals look so fake.

      • Re:That's nice. (Score:4)

        by GTRacer ( 234395 ) <gtracer308NO@SPAMyahoo.com> on Monday April 18, 2016 @03:57PM (#51935007) Homepage Journal
        Here's my datum. Saw it Saturday afternoon. Expected a lot of CGI. No glaring cases of cringey animation I can recall. Mouths seemed to be in sync.

        BUT - at no time did I expect 100% CGI. Several times I looked at an animal - mainly the wolves - and said to myself the bodies were likely real wolves, with the heads CGI-ed for acting purposes. The hair, musculature, movements, etc. were just... wolfy. When Bagheera (panther) chases Mowgli through the trees, I assumed most of the running action was a stunt animal, and the up-close conversation was CGI.

        I guess my "review" is that, unless you're there just to nit-pick, most everything just looks... natural. Oh, and the story? You already know it. But it's a fine implementation of the reference spec. A little more savage than I expected honestly!

    • Apparently it's good, according to critics and users?
      http://www.rottentomatoes.com/... [rottentomatoes.com]

    • Re:That's nice. (Score:4, Insightful)

      by Phics ( 934282 ) on Monday April 18, 2016 @12:43PM (#51933101)

      This isn't about the quality of the story, the human actors, or the script itself, it's about the tech being used to solve a problem. You can, on occasion, have a technical masterpiece which has nothing to do with the value of the actual project it was meant to accommodate.

      • Re: (Score:2)

        by g01d4 ( 888748 )

        Indeed, what takes 30-40 hours/frame to render has changed significantly since I was peripherally involved in the field almost thirty years ago. Rendering is one of the few areas that really need all the computing resources you can provide and fortunately it's able to leverage parallelism.

        I don't know how much cleverness is used -- things don't change that much in 1/24th of a second -- but I'm guessing there's still a lot of brute force computation. Directing becomes more critical in these efforts as the a

        • Re: (Score:1)

          by Jamu ( 852752 )
          If it can use parallelism, then the time per frame is set by making the film in a timely manner while spending as little money as possible. So the ratio of computer time to actual film time will be about 10k on average. Whereas 30-40 hours/frame is about 5,000k! I'm guessing that rate was per computer, and that's why they needed thousands of them. Was the entire film animated?

      • Didn't they try that with Avatar?

        Back then everyone decided that special effects be dammed, a big blockbuster movie does actually need a (non-plagiarised) story.

        • Re:That's nice. (Score:4, Insightful)

          by ShanghaiBill ( 739463 ) on Monday April 18, 2016 @01:38PM (#51933563)

          Didn't they try that with Avatar?

          Back then everyone decided that special effects be dammed, a big blockbuster movie does actually need a (non-plagiarised) story.

          No, the lesson of Avatar was exactly the opposite. The story was completely unoriginal (except for Pocahontas being ten feet tall and blue), but it grossed nearly $3 Billion, and was the most financially successful movie ever. That was purely because of the spectacular (for 2009) special effects. If the eye candy is good enough, the story doesn't matter.

          • There's a reason some call it "Eye Porn". Those things are never really lauded for their story either.

    • I hear the script only took one brain-dead writer a few hours to render.

    • Re: That's nice. (Score:1)

      by Anonymous Coward

      Better since they added the car chases and shoot-outs.

  • Literally (Score:1)

    by Anonymous Coward

    That required "literally thousands of computers,"

    Thank goodness it didn't require any figurative computers. I'm not sure how you'd get those.

    • A modern effects-heavy movie made with thousands of computers doing the rendering? Wow, that's totally unlike every other modern effects-heavy movie. Ever wondered why there are so many Dell computers in action movies with closeups of the Dell logo? Dell either loans, or gives a very steep discount on, the computers for the renderfarms in exchange for product placement.

      • Dell either loans, or gives a very steep discount on, the computers for the renderfarms in exchange for product placement.

        The cost of the computers is not much. The expensive part is the GPUs. An NVidia Tesla K80 costs $4000. The computer to host it costs only a tenth of that.

        • Right, and Hyperion doesn't (currently) use GPU acceleration. Extremely heavy SIMD acceleration, of course, but no GPU yet.

    • I guess that actually works if you count Virtual Machines as "Figurative" computers.

    • I'm not sure how you'd get those.

      It's called "docker".

    • Virtual machines, obviously.

  • Would a bear detect the uncanny valley? (Score:3, Interesting)

    by jellomizer ( 103300 ) on Monday April 18, 2016 @12:27PM (#51932959)

    Just curious. We as human are so intune to relate with humans, that a CG Human would look lifeless and fake, especially when animated. Non-Humans not so much so, as we don't relate on the same level. But would a bear watching it have a uncanny valley when seeing another bear?

    • Probably depends on the bear [youtube.com].

    • That's an interesting question - going by my experience with my dog and FaceTime, he doesn't have the ability to process a 2-D image as a representation of a living 3D thing. That's not a trivial thing - very young children can't do it either. Figuring out that a flat image at the wrong scale, with perspective distortion, lighting artifacts, reflections etc is equivalent to a real creature is (I think) a learned skill.

      When it comes to dogs (and maybe bears) I always assumes that partly because although we a

      • I've had one dog that would watch TV. She enjoyed dogs and horses in particular, as well as R2D2 due to his sounds. The 'evil' horses in Fellowship of the Ring got growled at. For the horses and dogs she would walk around to look behind the TV when they ran out of frame, when she was younger. When she was older she seemed to figure out they weren't really there but enjoyed it if there were dogs interacting with one another on the screen. Other dogs I've had were only interested in the sound the TV made
        • I had a dog once that would bark like mad and run to the front door when a doorbell rang on TV. It was amusing when I got a PVR to see how many times in a row she would fall for it.

      • Re: (Score:2)

        by swb ( 14022 )

        My dog will occasionally raise his head and look around if there's dog sounds on TV, but only if he's been sleeping or otherwise not alert.

        And maybe I just think it's the dog noises on TV, but in fact he's hearing something happening 4 houses away in the back yard that I can't here and the TV noise is just a coincidence.

        • Re: (Score:1)

          by ChoGGi ( 522069 )

          My buddy has a French Bulldog, loves people but a real bitch to other dogs.
          Any central barking sound from the TV she'd run up to it and start barking like mad, if the sound came from a side, she'd run to the window on that side and do the same.
          If it went on for more then a few seconds, she heads to the dog entrance and runs around the yard for a while (with a few barks).
          Some of the time she'd set off their other dogs and they'd all run around outside. This is on an acreage outside the city, so maybe it's ju

      • Also, the Red, Green and Blue components used to reproduce colors are finely tuned to match the absorption rates of the 3 types of cones found in the human eye.

        Simply speaking, the colors in our movies and photographs are probably quite unrealistic for most non-human life forms.

             

        • I think I read somewhere that the framerate might have something to do with it as well. Back in CRT days, you were really watching flickery still after still, whereas a modern TV gives a much permanent image, and in some cases it's automatically interpolated to be smoother, as well.

          Not sure if that makes a difference to the mogs and mutts.

    • Most animals aren't as visual as we are. For example, a dog's concept of "doggyness" comes from other dog's anal glands, if animals don't have that scent, they are considered non-dogs.
    • That might depend on what the TV smelled like.

    • But would a bear watching it have a uncanny valley when seeing another bear?

      It depends on two factor:

      - which sense does the considered animal use to perceive the world and other individual of the same specie?
      As mentioned, we human are quasi exclusively visual, with some auditive perception (voice) thrown in too, whereas other animal rely on other sense or other interpretations of sense. (e.g.: some snakes have infrared perception and use it to recognise other animal. They would not recognize a mouse if it were cold)

      - how is the social structure of the animal, how much does it need

    • Re: (Score:1)

      by Tablizer ( 95088 )

      But would a bear watching it have a uncanny valley when seeing another bear?

      Something tells me their brain will simply put it into one of 4 categories:

      1. Screw-able
      2. Delicious
      3. Hogging my territory
      4. Ignore

      "Real" versus "fake" is not something they can contemplate. Perhaps if something looks off, they may take it as a hint it could be injured or sick, but only in terms of the above categories.

      Hunting animals often are keen to notice sick/injured because those are the best targets. Thus, "moving funny" may

    • Unlike humans, animals don't all depend as heavily on their sense of sight to interpret the world around them. Witness that wild animals will attack a decoy that is not at all realistic-looking. I suspect that a live bear being confronted with a realistic CG bear would be more confused by the lack of smell than by any visual imperfections.

    • Re: (Score:2)

      by clovis ( 4684 )

      In addition to the possibility of uncanny valley problems when serving content to animals is the problem is that animals need a much faster fps to perceive continuous motion.
      Dogs need about 70 fps or faster or it looks like a series of static pictures much like 5 fps might look to us.
      Birds need over 100fps.
      Older TV's were probably annoying as all get out to them. I know my dogs from long ago disliked being forced to watch "The Smothers Brothers" with me. Now I know it was due to the fps being to low.

      Thanks

  • i want to identify as a man, man-cub

  • "So you can tell how much the computer has to figure out, exactly what it's doing, how it's bouncing, how much of the light is absorbed, because when it hits an object, some gets absorbed and some gets reflected."

    He's so in awe of this you probably shouldn't mention that reflected light can hit another object and keep reflecting... possibly mixing with light from a complete different source in the process.

    MIND == BLOWN

  • So we're, say, about five years away from no longer having any way of knowing what we're shown is true or total fiction...

    Then what?
  • all the characters look dry and boring compared to the racist original. the monkey who's supposed to be a cool black guy. baloo the irish drunk. the uptight british elephants. forgot who else

    • Re:cool CGI, but looks like it has no soul (Score:4, Informative)

      by TsuruchiBrian ( 2731979 ) on Monday April 18, 2016 @01:15PM (#51933385)
      The "monkey" was actually voiced by Louis Prima (a Sicillian) who was actually a pretty famous jazz musician and singer. He wrote and composed Sing Sing Sing, probably the most widely known swing song.

      • Re: (Score:1)

        by Anonymous Coward

        According to IMDB: [imdb.com] "Jazz singer Louis Armstrong was originally set to voice King Louie but another jazz singer Louis Prima was cast instead after Walt Disney feared that the idea of Armstrong who was African-American to play an ape would make the audience find the film racist."

        • That's true. But I still don't think that the King Louie that Louis Prima voiced was "supposed to be" a cool black guy, unless Louis Prima was trying to be a cool black guy, even though King Louie was originally "supposed to be a" cool black guy (Louis Armstrong), in a different sense of phrase "supposed to be".
  • Disney ran out of other people's intellectual property to rip off, so now they're ripping off their own? This movie will run into all the same traps that all movies which anthropmorphicize wild animals run into. My favorite is Zootopia: what exactly do all the carnivores eat, when ALL animals have evolved into hominids?

    • Re: (Score:2)

      by zlives ( 2009072 )

      humans

    • Disney ran out of other people's intellectual property to rip off, so now they're ripping off their own? This movie will run into all the same traps that all movies which anthropmorphicize wild animals run into.

      That trap would be the $1 Billon gross in first-run theatrical release for Zootopia? The 98% Fresh rating from Rotten Tomatoes?

  • Ball park numbers:
    35 hours per frame
    120 frames per second (2x 60 fps)
    7200 seconds (2 hours) per movie
    = 30,240,000 hours to render the movie
    = 3452 years

    I'm sure they have a cluster of computers dividing up the workload, but still...

    • They mentioned thousands of computers, and it's at 24fps not 60.

      • Film is traditionally 24 fps, but I a lot of stuff coming out now is 60, especially stuff shot in 3d. Even if some might want to view the film in 24 fps, "shooting" at a higher framerate gives you the option of a high framerate release.

        There are jungle book trailers that are 60 fps.

        • There are jungle book trailers that are 60 fps.

          No, there are people on YouTube who think they can run a 24fps film through an interpolating filter and get something that's on a par with native 60fps.

          You can't.

        • I'm sure that someone could take a clip of the trailer and stretch it out to 120fps if they really want to, but the movie was still actually produced using 24fps. That fact is in both the summary and the article. Why would they do 2 and a half times the work when they don't need to? Why should they spend 2 and half times as long to create the movie when they don't need to?

          • Why would they do 2 and a half times the work when they don't need to? Why should they spend 2 and half times as long to create the movie when they don't need to?

            They don't need to do anything this. They don't need to make it so realistic. They don't need to make a 3d version. They don't even need to make a movie. Why would they do any of this?

            • Yeah, exactly, fantastic reasoning. They should just produce everything at 1,200 fps for that matter, right? Let's really future-proof what we're doing and just set 12,000 fps as the standard. Because if people can really only detect 24 fps, who cares, why not just do 500 times the work to produce a product that people will experience the same way?

              Assuming that neither of us are completely retarded, if they have an expected workload measured in tens of millions of computing hours, why are they going to i

              • Because if people can really only detect 24 fps, who cares, why not just do 500 times the work to produce a product that people will experience the same way?

                Are you blind?

                VR headsets are trying to bump up the standard video game rendering rate to 60 to 90 so people don't get motion sickness

                If you don't think people can tell the difference between 24 fps and higher frame rates, you must have some pretty severe vision problems.

                Yeah there are frame rates high enough where human beings can't differentiate them anymore, it's just not anywhere near 24.

                • And, therefore, Jon Favreau decided to make The Jungle Book using 60 fps? Are you still after that claim? You should rework your ballpark numbers. You should also use a time of 3 hours for the movie, because I just think that story should be told in 3 hours regardless of what the reality is. If you're not going to bother to use the actual time or the actual frame rate then might as well make it so that it just feels right.

                  • And, therefore, Jon Favreau decided to make The Jungle Book using 60 fps? Are you still after that claim?

                    I *never* made that claim.

                    In fact I specifically said "Ball park numbers" (i.e. I am guessing).

                    The claims I am making are regarding the utility of shooting in higher frame rates, and the history of modern movies (especially those focusing heavily on VFX/CGI) being shot/rendered in higher frame rates.

                    Even if a director wants to present their movie in 2d 24 fps film, there may be a market for IMAX 4k 3d 60 fps, blah blah blah version. You can always downgrade form higher quality source material.

                    You should also use a time of 3 hours for the movie, because I just think that story should be told in 3 hours regardless of what the reality is.

                    If most curr

                    • If most currently produced movies where 3 hours long, I probably would have used that ball park number.

                      That's my point - if your goal is to estimate the total amount of computer hours used to render this movie, then why use anything other than the actual duration (105 minutes) and the actual frame rate (24)? Seriously, why use anything else? It's only going to be an estimation anyway, why introduce additional error into it?

                      You had this:

                      35 hours per frame
                      120 frames per second (2x 60 fps)
                      7200 seconds (2 hours) per movie
                      = 30,240,000 hours to render the movie
                      = 3452 years

                      Let's adjust for reality:

                      35 hours per frame
                      48 frames per second (2x 24 fps)
                      6300 seconds (105 minutes) per movie
                      = 10,584,000 hours to render the movie
                      = 1208 years

                      There you go

                    • That's my point - if your goal is to estimate the total amount of computer hours used to render this movie, then why use anything other than the actual duration (105 minutes) and the actual frame rate (24)? Seriously, why use anything else? It's only going to be an estimation anyway, why introduce additional error into it?

                      Do you not know what guessing is? Does it ever make sense to guess when you know the answer to something?

                      There you go, still a ball park estimate but up to 285% more accurate!

                      except that you don't actually know how much more accurate it is because you are also guessing.

                      Is that where I said this: "Because if people can really only detect 24 fps"

                      yes

                      Because, based on words like "can really only" or "most people", I'll go with an answer like this:

                      The difference being that I have admitted from the beginning the things I am guessing about, and you've apparently just started now

                      That's a weasel way out. My original reply to you only said that the film was done at 24fps instead of 60, and you responded with how a lot of things coming out now are 60, and that you've seen trailers for the movie also at 60 (why does any of that matter if this movie was done at 24?).

                      It matters because I am suggesting that it's possible the movie will actually be rendered at a higher framerate than 24 even though that specific number was casually mentioned i

                    • except that you don't actually know how much more accurate it is because you are also guessing.

                      The only guess is the average number of hours per frame. The time in seconds has a margin of error of +/- 59 maybe? Assuming that they would truncate instead of round when reporting the duration. Your numbers guess at all 3 of them, even when a guess is not even necessary because we know.

                      And even with your "more accurate" numbers, my ball park number is still within an order of magnitude (i.e. ball parks are large).

                      Congratulations, your estimate is probably accurate within a factor of 10x.

                      I think the real issue is that I can't post some *ball park* numbers with a disclaimer, without someone trying to nitpick it.

                      Welcome to Slashdot.

                    • The only guess is the average number of hours per frame. The time in seconds has a margin of error of +/- 59 maybe? Assuming that they would truncate instead of round when reporting the duration. Your numbers guess at all 3 of them, even when a guess is not even necessary because we know.

                      This was a criticism of the accuracy of your accuracy number.

                      Congratulations, your estimate is probably accurate within a factor of 10x.

                      According to you it was was within a factor of 2.85x. So the prediction appears to have exceeded expectations.

                      Welcome to Slashdot.

                      I am under no illusions that people like you are uncommon on slashdot.

    • Imagine if all that computing power went into mining bitcoins instead...

  • Slow rendering is a good thing now? (Score:3)

    by GuB-42 ( 2483988 ) on Monday April 18, 2016 @01:38PM (#51933571)

    Making a slow but accurate rendering engine is easy. The equations are relatively simple, the rest is simply a lot of iterations on crazy detailed model.
    Making a fast but good looking and sufficiently accurate engine is fucking hard.

    From the summary, it looks like the team went with the bruteforce approach, which is fine, but not something to brag about IMHO. They even imply that it takes twice as long to render from two very close viewpoints (stereo). I'm surprised they can't manage to exploit the correlation between the two.

  • No matter how much computing power they throw at rendering the result has always sucked.

  • Yawn (Score:2)

    by wcrowe ( 94389 )

    Yawn. All that effort, and yet it's still the same basic story as fifty years ago. Lots of tech... zero creativity.

    • Re: (Score:2)

      by LWATCDR ( 28044 )

      You left out that Disney waited for the Jungle Book to enter public domain before they made the first movie.... Under the old rules of course.

      • Re: (Score:2)

        by wcrowe ( 94389 )

        Indeed. I'm not sure Disney has had a truly original concept since Steamboat Wille. Disney, an entire entertainment empire built on a single cartoon.

        • Re: (Score:1)

          by tbq ( 874261 )

          Indeed. I'm not sure Disney has had a truly original concept since Steamboat Wille. Disney, an entire entertainment empire built on a single cartoon.

          Steamboat Willie was based heavily on the Buster Keaton movie Steamboat Bill Jr. which came out earlier that year. It was the first Mickey Mouse cartoon to get a distributor, since Disney couldn't find a distributor for the two earlier Mickey Mouse cartoons.

  • As powerful as the computers were, they ultimately were just taking cues from the human innovators..."there's no real computer that replaces the skill of the...person who is pushing the buttons."

    Yeah, that's what H-1B's are for. [cnn.com]

  • > How 'The Jungle Book' Made Its Animals Look

    The trailers I've seen look terrible. As is too often the case, the kinematics are just *wrong*.

Slashdot Top Deals

It is better to travel hopefully than to fly Continental.

Close