Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
Graphics Technology

Newly Digitized Film Shows Ed Catmull's 3D Graphics From 1972 95

AlejoHausner writes "In 1972, Ed Catmull, then at the University of Utah, put together a film showcasing many of the 3D computer graphics techniques he and others had developed while working as students in Ivan Sutherland's lab. That film has been digitized and is available. All kinds of modern techniques like Gouraud shading, deformed meshes, and z-buffering are shown in the film. There is a segment showing Catmull digitizing a plaster model of his hand. Catmull later founded Pixar, but at the time the Utah lab pioneered many of the graphics techniques we take for granted today." I'm just sorry I missed when this film was first made available online earlier this year.
This discussion has been archived. No new comments can be posted.

Newly Digitized Film Shows Ed Catmull's 3D Graphics From 1972

Comments Filter:
  • Thanks for that link, as a graphics artist - this brightened up my day!

  • by lkcl ( 517947 ) <lkcl@lkcl.net> on Saturday September 03, 2011 @04:58AM (#37295160) Homepage

    i wonder how much this buggers up any companies doing patents on 3D GPUs? the reason i ask is this: one of the problems that the ARM embedded SoC vendors face is that they are stuck on choice for GPUs, from companies who have had to design very low-power 3D engines (Vivante etc). these companies are quite young, and their relationship with the "big boys", who have had over a decade to establish their "arms-race" arsenal of patents, is unclear. so the embedded SoC 3D companies are LESS likely to release free software drivers. but if the very foundation of key parts of 3D patents is undermined through prior art.... i dunno...

    • ... but if the very foundation of key parts of 3D patents is undermined through prior art.... i dunno...

      Formulas can't be patented, so it's unlikely that this video could provide any prior art for dismantling patents. Probably about all this video and the GPU patents share in common is the formulas involve.

      • Formulas can't be patented,

        Not sure that is accurate. Written as an algebraic expression, perhaps. Expressed as a method or a device (ie: drug) then a formula is able to be patented. Formulas as software (as pointed out by AC) is another avenue, such as MPEG. A formula is simply a mathematical expression, which describes a lot of software that is now patented, and as it stands now, software can be patented, at least in the US.

        • Formulas can't be patented,

          Not sure that is accurate. Written as an algebraic expression, perhaps. Expressed as a method or a device (ie: drug) then a formula is able to be patented. Formulas as software (as pointed out by AC) is another avenue, such as MPEG. A formula is simply a mathematical expression, which describes a lot of software that is now patented, and as it stands now, software can be patented, at least in the US.

          Software contains flow control, conditionals, a process, formula describe mathematical relationships and values but are not process. We have regularized processes for evaluating them, however they are not "process" in and of themselves.

      • ... but if the very foundation of key parts of 3D patents is undermined through prior art.... i dunno...

        Formulas can't be patented, so it's unlikely that this video could provide any prior art for dismantling patents. Probably about all this video and the GPU patents share in common is the formulas involve.

        In theory the law says that they can't in practice patent offices do approve patents for them and going to court to void them is a roulette.

        • ... but if the very foundation of key parts of 3D patents is undermined through prior art.... i dunno...

          Formulas can't be patented, so it's unlikely that this video could provide any prior art for dismantling patents. Probably about all this video and the GPU patents share in common is the formulas involve.

          In theory the law says that they can't in practice patent offices do approve patents for them and going to court to void them is a roulette.

          You've responded with the best response to my comment. Patents are often issued for things that are not strictly patentable, but taking them down is of course, a wonderfully painful legal matter.

    • by gl4ss ( 559668 )

      the algo's for basic 3d are now pretty old and much of it is non-patentable, it's some hardware implementation patents which matter for the manufacturers more probably. and some other things like s3's texture compression but you could find prior art that does exactly almost the same(just not with a non-reprogrammable hw part). btw one reason why embedded 3d providers have been reluctant to release oss drivers is that some chips just don't do what they advertise.. and that the drivers have code they've licen

    • by Rockoon ( 1252108 ) on Saturday September 03, 2011 @07:15AM (#37295428)
      GPU's dont use the lighting techniques seen here, in spite of the fact that the summary claims that the technique is "modern."

      Specifically, nobody does gouraud shading anymore. Hell, even in the days before GPU's people stopped using that technique in favor of phong shading.

      Gouraud calculates the lighting at each vertex and then interpolates the light intensity across the polygon.
      Phong calculate the surface normal at each vertex and then interpolates the normals across the polygon (calculating light intensities from those normals on a per-pixel basis.)

      Hell, nobody does phong any more either. Generally the normal is now either a direct lookup (bump mapping and so forth) or derived from the zbuffer itself using differed shading.
    • by KliX ( 164895 )

      None what-so-ever. This stuff is all published - it's Catmull for god's sake!

    • by mikael ( 484 )

      Back then, the patents related to hardware implementations for lighting calculations. Those patents would have expired.
      Basic 3D API calls are just to draw lines, fill triangles using texture mapping. There isn't anything to patent there now. Even those low-power GPU's support programmable shading models.

      If there is anything worth patenting, it will be related to parallel processing at the ASIC level and advanced lighting models at the mathematics level.

      ID Software did get threatened with a lawsuit from Crea

  • Didn't the wireframe animation appear in a monitor in a scene in "Westworld"?
    • Didn't the wireframe animation appear in a monitor in a scene in "Westworld"?

      Indeed, the TFA mentions this very fact.

      • Futureworld [wikipedia.org].
        • And apparently TFA doesn't even mention this at all... now I look like a total idiot...

          • by amiga3D ( 567632 )

            Nah! You noticed you were wrong. The total idiots never figure that out.

            • Nah! You noticed you were wrong. The total idiots never figure that out.

              Ok, so just an idiot then. :)

          • TFA may not mention it, but TFF does - it's in the introduction of the film.
            • TFA may not mention it, but TFF does - it's in the introduction of the film.

              If the article contains the film, then does it count? I could have sworn I had seen it somewhere in the article... silly me, forgetting that there was text in the video itself. >_

          • by CityZen ( 464761 )

            The lead-in for the video itself mentions Futureworld, which is probably what you were thinking of.

    • by KalvinB ( 205500 )

      If you watch the documentary about Pixar (available streamed on Netflix) it shows the scene and talks a bit about the hand animation. This video actually shows a lot more though,

  • by Anonymous Coward

    Charles Csuri has a video like this from 1969. First NSF grant to an artist. Look it up. This is impressive but it's not the first. Csuri's students and coworkers went on to found Pixar. Note: I work with him and also have met Sutherland, et al. when I worked on the DARPA HPCS project at Sun.

    • by suso ( 153703 ) *

      And Eric Graham (maker of the juggler animation) beat them all to it. He was making raytraced 3d graphics in the mid 1960s on teletype output. Actually, I find this whole video's claims of being the first time it was digitized a hard to believe. I've seen clips from this video before and for as important as this video is, it would have been digitized in the last 20 years instead of just left on a shelf to chance be picked up by someone's son like a family home movie.

    • by CityZen ( 464761 )

      Csuri's work is certainly 3D computer graphics that predates Catmull's film, but it's the particular techniques shown here that make Catmull's film remarkable. Csuri's work from this period (that I've seen) is only rendered points and wireframes without any hidden-surface removal. Catmull shows fully shaded polygons with correct depth ordering (it's likely he used techniques other than Z-buffering to achieve this).

      The founders of Pixar are Ed Catmull and Alvy Ray Smith. See: http://alvyray.com/pixar/defa [alvyray.com]

  • by tehdaemon ( 753808 ) on Saturday September 03, 2011 @05:54AM (#37295292)

    Pixar - A Human Story of Computer Animation [youtube.com]

    If anything, this video is too long, but it gives a lot of background on Ed Catmull , the animated hand, and Pixar. Well worth the time, especially if you don't know what's the big deal with a crappy hand animation.

    For example this video was probably made by taping polaroids to a CRT to get the images out of the computer and onto the film.

    T

    • For example this video was probably made by taping polaroids to a CRT to get the images out of the computer and onto the film.

      I haven't had time to watch the whole 100 minute video yet, but are you talking from having seen this film or just guessing?

      My assumption had been that they would point a camera at a display, possibly cycling through multiple renderings of the same image using different dithering randomisation settings so that if (say) they only had a 2-color (black/white) hi-res display they would be able to "average out" and simulate greyscale. Of course, even a single 320 x 192 two-colour image would take up 8 KB, whic

      • Not guessing - just poor memory - it had been months since I had watched it. Ed does talk about a polaroid camera, seconds or minutes per frame, and implies 3 exposures per frame for color pictures. Chances are the image was rendered one line or pixel at a time. Nobody had a real framebuffer yet, publicly anyway. They mention that too.

        T

    • Wow, neat. I was working on a project once to automate fitting of endographic stents and in my CG book there was a section on Catmull-Rom splines, which fit the bill (and the blood vessel) perfectly. I hadn't realized there was so much to the back story.

  • I for one am glad that they didn't think to themselves "hey - I wonder if there's money to be made in the adult entertainment industry with this stuff..."

    Seriously... it must have almost crossed their mind... and then we would have been without Toy Story and a whole load of other great films!

    • I for one am glad that they didn't think to themselves "hey - I wonder if there's money to be made in the adult entertainment industry with this stuff..."

      Almost forty years on and computer-generated porn still sucks (and not in the desired sense either). Creepy, mannequin-like, uncanny-valley figures in crappy fake poses. Granted, I may be biased because I like the more naturalistic stuff, but it's still rotten.

      Seriously... it must have almost crossed their mind... and then we would have been without Toy Story and a whole load of other great films!

      Well, unless the bad publicity would have precluded them from making Toy Story, why couldn't they have done both?

      • by jamesh ( 87723 )

        Seriously... it must have almost crossed their mind... and then we would have been without Toy Story and a whole load of other great films!

        Well, unless the bad publicity would have precluded them from making Toy Story, why couldn't they have done both?

        I might hesitate to take the kids out to see Toy Story if the poster contained something like "From the makers of Sarah meets the Tentacle Monster from the planet Eroticon 6". Also, Pixar love including subtle references to their other films... you just never know what they'd manage to sneak past the censors :)

        On the other hand, i'm sure the "blooper reel" would be hilarious on their adult films!

  • What trick did they use to animate the wireframe ? If it took them a 3d manual digitizer to recreate the 3D model, how did they animate fingers ? Did they digitize every frame ? Were they already using skeletal animation ?
    • The model hand is rigid. Letters in the 3D titles morph into each other. So my guess is: Based on the digitized geometry, they manually created the geometry for key animation phases and interpolated the rest of the geometry.
  • I remember seeing the faces (5'10" onwards) and some other bits on some BBC or Channel 4 TV documentary a while back now (late 90s) and being very impressed by the fact they'd been able to do that in 1972 (matter of fact, I'm more impressed with that than I was with the hand). In terms of quality, I would have guessed it was done years later, more like the late-70s.

    Having looked for it online more recently (on YouTube and via Google) I wasn't able to find it anywhere, so it's great that these are now avai
    • by jovius ( 974690 )

      I happened to listen to early Jean Michel Jarre music while watching the video, and it was really fitting. I bet this piece ends up to the playlists of VJ's around the world.

    • by mikael ( 484 )

      Was that BBC Horizon - Painting with Numbers? That was my all time documentary back then. Horizon's intro sequence was art itself.

      That documentary showed some fascinating animations - wireframe drive-through of a down-town area, terrain fly-overs, a textured cube with some animated textures (Sunstone?), and a light pen system which demonstrated cartoon animation - a little cartoon character had a nose that expanded balloon style and pulled him upwards.
      There was a African-American kid who did some sprite pro

      • Was that BBC Horizon - Painting with Numbers? That was my all time documentary back then. [..] There was a African-American kid who did some sprite programming with a TI system.

        If the reference was to one of TI's computer, and it was a contemporary example- suggesting that the documentary was made in the early '80s- then no, it wasn't that one. The one I saw would have been around the mid-to-late 90s.

        Sounds like it- and the other stuff you mention- might be worth checking out via official or, er, unofficial channels!

  • That was truly amazing. I'm impressed, and thanks for sharing that.

    How, though, was it output? Obviously what we saw was a digitized version of film, but how was the film made in the first place? As a kid in the late 1970's, microcomputers (that's what we called PC's then) already output modulated signals which could be recorded on early VCR's. How were these put into film? And was it real time, or generated frame by frame a la Pixar?

    • by dr_dank ( 472072 )

      I didn't think any computer back in 1972 had any kind of graphics capability outside of ASCII art.

      • Computers in the 60's not only had graphics, they had mice! http://www.youtube.com/watch?v=1MPJZ6M52dI [youtube.com]
      • by mikael ( 484 )

        History of Framebuffers [wikipedia.org]

        In 1969, Joan Miller of Bell Labs experimented with the first known instance of a framebuffer. The device displayed an image with a color depth of three bits.

        In 1972, Richard Shoup developed the SuperPaint system at Xerox PARC. This system had 311,040 bytes of memory and was capable of storing 640 by 480 pixels of data with 8 bits of color depth. The memory was scattered across 16 circuit boards,

        In 1974 Evans & Sutherland released the first commercial framebuffer, costing about $1

    • Frame by frame, probably, though wireframe could surely be rendered realtime or near that.

      As for filming, camera rigged to a CRT and some electromechanics to control shutter/film feed after rendering the current frame could do the trick

    • My guess would be that they used grayscale graphic display + triple on film exposure of each frame using color filters.
    • by CityZen ( 464761 )

      I was wondering that same thing. I'd guess they made a film transfer recorder using a computer-controlled oscilloscope. This is essentially a film camera focused on an oscilloscope display (in a light-proof enclosure), where the computer can vary the position and intensity of the light spot (using D/A converters driving scope inputs). If you use a white phosphor CRT and some color transparencies in between, you can do full-color recording. (This is exactly the same principle used in some modern film rec

      • by tibit ( 1762298 )

        An oscilloscope display is electrostatically deflected. You cannot get any decent size out of the image that way without making it huge (long) or it being very slow, pick your poison. You need the image to be large since the spot is nowhere near "tiny". Modern oscilloscope tubes use mesh expanders and those diffuse the spot.

        I think they'd have used an electromagnetically deflected monochrome CRT with a phosphor that gives enough light in red, green and blue. They would have tweaked the geometry (in the CRT

        • by CityZen ( 464761 )

          Thanks for the extra insights. I wasn't aware of the internal differences between oscilloscope CRTs and TV-type CRTs. I only knew that they have X & Y (deflection) inputs and a Z (intensity) input, ideal for hooking up to a computer's D/A outputs, back in the day.

          I was aware of laser film scanners, but the only one I've seen (at Pixar, in fact) was quite a big, ungainly-looking device. I have no idea how common they might be, outside of movie-making houses.

  • by tverbeek ( 457094 ) on Saturday September 03, 2011 @08:23AM (#37295646) Homepage

    It may not be as technologically advanced, but it has a better plot than [insert recent digitally-rendered feature film here]!

    • Actually what I love most about this film is how the B&W rendering and the use of intertitles and accompanying music, gives it the flavor of a silent movie, making it feel even older than it is.

      • Well, it actually is a silent movie - the blog author says that he had the 8mm reel digitised, and added the music (apparently by Dave Brubeck) himself.
  • I never knew 3D computer graphics were developed back in the 70's. Interesting video buts its lacking any technical explanation of how they did it.

    What computer was this developed on? What programming languages were used? Was there a need to develop any special hardware to enable this rendering? The animation looks like its running in real time, where computers of 1972 fast enough to handle this? Or was this filmed by rendering one frame at a time and exposing a single frame of film? The computers of those

    • I agree that the like of detail is disappointing. However...

      The animation looks like its running in real time, where computers of 1972 fast enough to handle this?

      I think I could safely say that if I knew *nothing* about that film beyond it having been made in 1972, that I'd still bet my life on it *not* having been done in real time! Seriously, I'm pretty damn sure that you couldn't have done that in anything approaching real time even in 1982...

      • by blair1q ( 305137 )

        Nope. Not a chance. Catmull worked out the math. He had no clue about 3-D hardware because until he'd shown the rendering in software nobody else knew what it should do.

        Real-time, if it wasn't just dumping bits to a file for playback, you could probably see this thing drawing the wire frame line-by-line and filling each pixel.

    • I don't see how that could have been rendered in real time, especially the fully shaded hand. Think of all those pixels! Each frame was rendered on a CRT and photographed individually.
    • It's impossible even for the wire frame version to be rendered in real time. Hell for those computers it was impossible even to display in realtime the pre-rendered frames. The movie has been assembled by stitching together photos of individual frames on a computer screen.
      • Nope, wireframe would be easy - vector displays were widespread, so rendering wireframe is just transforming and clipping lines from 3d to 2d projection and sending points to the display.

        Still, it probably didn't make for smooth real-time animation, but should be enough for interactive use - editing/previewing/etc.

    • by Gumber ( 17306 ) on Saturday September 03, 2011 @04:58PM (#37298606) Homepage

      I dug into the technical details a bit and posted some of what I found on my blog, along with links to the papers describing the hand and facial animation work in more detail: http://geekfun.com/2011/09/03/early-cgi-animation-by-ed-catmull/ [geekfun.com]

      The short answer is that the facial animation was produced by software written in Fortran and run on a pair of PDP-10s, and the hand animation was likely running in the same environment. When each frame was finished, it was displayed on a CRT and captured to film using a 35mm animation camera. For the facial animation, each frame took about 2.5 minutes to render.

      • by CityZen ( 464761 )

        Just to be clear: lacking a framebuffer, the system could not display images in such a way that you could view them with your eyes (unless you used a storage CRT http://en.wikipedia.org/wiki/Storage_CRT [wikipedia.org] ). Rather, it would scan out a single dot at a time while the shutter on the camera was held open. After scanning the 1024x1024 array of pixels (which took about 2.5 minutes), the shutter would be closed, and then the film could be developed, and only after it came back from the film lab would you see your

        • by Gumber ( 17306 )

          Thanks for the clarification. I'd misread the 2.5 minute time as being the total throughput, not the time it took to output a single completed frame, but rereading the paper, it seems like it is indeed the time to expose a 1024x1024 frame. Its unclear to me how long the computation took.

  • Looks better than Crysis 2.
  • The model (of a hand) was "digitized" - get it?

So... did you ever wonder, do garbagemen take showers before they go to work?

Working...