Back to Index

Special Report: NAB 2001

9th May 2001

More NAB 2001 Reportage: Video Products from Anon...


(The original Lightwave report is here)

3DV

The most innovative product I saw was by an Israeli company called 3DV, where they were using an infrared system mounted on a video camera to capture Z-depth data along with video at the same time so that they could key out the subject based on z-depth information instead of the traditional use of blue-screen. Of course, the information could be digitized as well, so that compositing software could use the z-depth information for depth-sensitive effects. I don't think the resolution was quite there for film-level quality, but I thought the idea was ingenious. They were able to key a guy out from a busy background of displays and moving people without any trouble at all. Imagine how that would work with a steadicam on a location unencombered by big sheets of blue-painted wood?

Realviz

RealViz caught my attention for sheer power of products. They have four products that they were demoing: Stitcher (for making Cubic Quicktime VRs), ReTimer (for creating film/video frames that are interpolations of existing frames so that they can create footage that's the equivalent of shooting with a 300 fps camera), Image Modeler (which lets you create vertices of an object from multiple pictures of that object, from which you then create a mesh and fully textured, animatable 3d model), and MatchMover (motion analysis of camera path/zoom for export into 3D applications).

First, I'd like to point out that they intend to have all products running on the Mac (OSX) by next year.

One of their demo artists commented on the number of people who had asked him about Macintosh support. Not bad for only making up 5 of the computer market, eh?

Second, their applications all share an incredible pixel identification engine as they can track pixels as obscure as the line between lip and face skin at incredible speed. Whether aligning pictures for a QT VR (which I will use After Effects to do, since I have yet to see another program give me that kind of control before now) or tracking 15 points in MatchMover, their system of analyzing images is beyond Commotion's for instance.

With Stitcher, you start putting pictures into a grid and as you adjust the pictures, the computer begins calculating what focal length your camera was using, and the more pictures you adjust, the more accurate its calculations until it's able to assemble the entire collection in proper perspective, analyze the overlapping edges to make sure pixels match, and then apply a cross fade that eliminates any seams. I'm very picky about such things, as I said, and this program made what I do in a day happen in 15 minutes. A truly intelligent program. It also ouputs to Pulse and Cult3D.

MatchMover is a special effects dream come true. I had the demo guy do a bit with some hand-held video that involved a zoom, and within 10 minutes, the 30 second clip was ready to be exported to my favorite 3D application so that I could embed objects to my heart's content, and only have lighting issues to deal with. The motion tracking was incredibly accurate and processed each frame in less than a second while tracking multiple points.

When I set Commotion Pro to its highest settings I never get the same level of accuracy they were getting, and even with Commotion taking advantage of the G4's altivec engine, it's nowhere near as fast. MatchMover worked so fast that you always felt you were working -- no need to take coffee breaks while waiting for it to track a single point. Image Modeler was like Canoma but far better. Instead of letting you just create basic geometric shapes from a photograph or two, you could model someone's head from a slew of photographs, and get incredible accuracy. Using it's camera analysis and pixel-tracking capability, it could figure out where a camera was when a picture of an object was taken. Each time a pixel in one photograph was defined by the user, it would then identify the same pixel in other photographs, and thus build a 3D cloud of points. From these points, a mesh of polygons could then be created. Points and meshes could be manipulated for greater accuracy. Then the program would map the mesh with textures taken from the photographs that were taken at positions most normal to the polygon being mapped. When the computer noted adjacent polygons being mapped by a different photograph, it would blend the joining seams to eliminate unwanted differences. At Siggraph they'll be showing an animated woman that was modeled entirely in Image Modeler by having a lot of photographs taken of her. I got to see the model of her head, and it was great under OpenGL!

Obviously, none of these packages are to the point of just pressing a button and walking out the door -- having the skills to track objects or model are what keep these programs inline, but they certainly do open up the horizon a little.

Finally, there was ReTimer. As I was watching the Image Modeler guy, someone slipped him a black business card and told him to let them know about any future products they had in development. I looked at the card and it said "Industrial Light and Magic." I remarked to the demo guy about that and he said that ILM had just bought 10 copies of ReTimer while there at the show! ReTimer is the incredible software program that lets you adjust the frame rate of existing footage through vector analysis and frame recreation. I won't go into the demo procedure, but they could take footage of horses running toward the camera and slow it down to 1/10th it's normal speed, and make it look absolutely perfect -- the mane waving in the wind, the legs working in perfect harmony, and no jitter or blur to cover up any bad spots. I thought it looked perfect. They showed footage of a couple of Orcas jumping out of the water at Sea World against big BlueScreen backgrounds with water flying off their bodies in the glistening sunlight, and water splashing as they reentered the aquarium. Then they slowed it down to 1/10th and the water still glistened like jewels, and the Orcas still flew -- only slower. Absolutely jaw-dropping, tongue-hanging stuff. They could force already shot footage to vary in frame-rate to duplicate the commercials we've all seen shot with expensive high-speed cameras. Truly, the computer is becoming the average-income guy's best friend (OK, maybe the average guy would have to sell his only car to afford all this, but at least it's a choice now :)

Re:Vision

A stop by Re:Vision's little stand, and lo and behold they have twixtor -- an AfterEffects plug-in that does what ReTimer does without all the controls that ReTimer has (and at 1/7th the price for the Pro version -- around $500). I just tried their demo out on some footage of my son when he was toddling around at 14 months, and was able to slow him down to 1/10th speed with remarkable clarity. There was some picture crawl when he moved his arm past a picture on his shirt, but Re:Vision is working on yet another piece of software that eliminates video interlacing so completely that it looks like film -- better than what I've seen Photoshop or After Effects do, for sure. You might want to check them out at www.revisionfx.com. I've been using their motion blur filter for over a year now to save my butt when I don't have time to do a motion blurred render in LightWave but want to avoid the strobed look when I export a composite for video. They do really good work, and their plug-ins are remarkably affordable for their performance.

Sorensen/Quicktime

A guy from Kamera Technology (Declan Caulfield) spoke at Sorenson's booth about Quicktime. He was pretty stoked about it, claiming it was the only streaming solution out there that had built-in interactivity. Evidently, his group had been tasked with setting up streaming services for last year's Olympic Games, and he used Quicktime and Sorenson's (beta) MPEG-4 compressor. He showed how he had set up Quicktime to stream sporting events for his home country of Sweden, with Picture-in-Picture movies playing at the bottom of the Quicktime player, a good-sized movie playing above and to the right, and a menu (with chat capabilities) on the upper left. He said that once viewers of the games were able to chat and watch at the same time, average viewership went from 5 minutes to over an hour. He also said that viewership also became community-driven as someone might type "Oh, check out channel 3, so-and-so just fell off his horse" and everyone would switch to channel 3 to see what was happening! And this was with Quicktime 4! Quicktime 5 is slated not only for MPEG-4, but Apple (as I'm sure you know) has also promised MPEG-2 support as well. My only complaint is that they don't make it more obvious to use all it's capabilities. However, I do think they're on the right track, because MPEG-4 is more than just about video compression -- it even supports 3D and picture formats to a worldwide standard of visual communication by the MPEG committee.

Electric Image (Alex Lindsey - DV Garage)

Finally, I had to visit the Electric Image booth. Alex Lindsey (they guy who surfaced the queen's ship in Star Wars Episode 1 and who also modelled and animated all the ice crystals in Titan AE) was demoing Universe there. He also is the founder of www.dvgarage.com which I like to frequent for his insight. He was promoting Universe's layered rendering system, and I was amazed at the level of detail he'd go into. It wasn't enough to just generate a map for each surface type. He'd generate multiple maps for specularity and reflection, for instance. One with a broad reach across objects, and a second with a more confined and more pronounced effect. All these layers were then tamed with grunge maps, and combined to create a level of realism that would sink Bill Flemming :) (OK, I've never read his photoreal book, but a little dig at the guy who conned a Serious 3D subscription from me is in order.)

I'm always amazed by guys who've reached the top (ie. working at ILM, or doing work that makes it into serious-budget movies) and manage to be personable and friendly. We also talked briefly about his undertaking to train Africans in various nations on 3D modelling, starting with Amorphium Pro, in an effort to help them skip the industrial revolution right into the current computer age. They are rich in artistic heritage, and he doesn't see the point in continuing to give third-world countries our hand-me-down trades when we can all benefit by giving them more. I liked that guy.

After Effects 5

I did sit in on two After Effects presentations at NAB as well (partly motivated at catching a flying T-Shirt, but intrigued by the demonstration nonetheless). It is a significant upgrade, as there is now a switch for turning a layer into a 3D enabled layer. You get to add lights with adjustable properties, and camera(s) with adjustable properties, and your 3D-enabled layers suddenly gain some Lightwave-like properties (diffusion, specularity, shadow-casting on or off, shadow-receiving on or off, etc.) By adding, say, 10 layers of video, each one can be made to fly past the camera in 3D space at different distances and speeds to create a cool effect that previously could only be done in a 3D application by mapping video (or, worse, video converted to stillframe sequences) to flat polygons that were then flown past a camera).

Then, just when you thought that was good enough for an upgrade, the demo-artist breaks through your imagination by using the phrase "point-and-shoot." Any Mac-head knows "point-and-click" and "plug-and-play" as affection-laden phrases, used the first day we touched our Macs. But "point-and-shoot" is to LW's expressions what elegance is to bare-knuckled survival :) With "point-and-shoot" you can simply draw a line from one layer to another to create a parent-child relationship. Or, you can enable a layer to support expressions (basically, javascript), and then "point-and-shoot" to any properties on any other layer that you want your expressions to be based on.

As an example, the demo-artist typed out the word "IMAGINE" (one letter per layer), and then spun the letter "G" around once. He then attached expressions that were dependent on the G's rotation to the other letters. The other letters then spun around a vertical axis (centered on the center of the screen) in different speeds and directions with only two keyframes in the entire composition -- the start and end keyframes for the letter "G"s rotation. He then raised the camera on the y-axis, rotated it down to look at the spinning letters, gave each letter a random starting point, and then had the camera drop into it's final position as all the letters swirled around to finally spell the word "IMAGINE" as the camera settled.

Another demonstration was just on the parent-child relationships that could be set up. Using Illustrator artwork that had been shaded with a nice 3D feel, (3 being organic-metal arms, the 4th, an organic-metal video monitor), he was able to parent each layer to the one below it, and then rotate the objects (after adjusting their pivot points) to give a cool animation of a multi-jointed mechanical arm retracting a spinning metallic monitor. By making adjustments to the simulated shutter-angle (from the default 180 degrees to 720), he was able to introduce beautiful motion blur despite the relatively slow motion of the animation. It looked like lots of fun.

Well, that about sums up my highlights. I did visit other booths that were offering streaming video services on the PC platform, digging for someone who could stream and offer interactivity (using email doesn't count with me). Pinnacle had a $20,000 broadcasting package that would let you stream video and Powerpoint pictures side-by-side. I had to chuckle when I saw a Cleaner 5 demo taking place in the Microsoft Partners pavillion. In the upper left corner of the screen was a tell-tale Apple logo. In a previous demonstration there, somebody had been trying to demo Premiere and couldn't get the sound to work and it didn't have any Apple logo in the upper left corner either. Evidently, a computer had to be swapped out to keep the Microsoft Partners crowd impressed :) I'll bet they never noticed the difference . . .

Apple's presence there was outstanding, too. Crowds were watching the FinalCut Pro demos and DVD Studio Pro demos. They must have had 40 G4's demoing everything from Maya and Discreet stuff to Quicktime's media skins and just plain old OS X -- I guess it'll be at least another year before Apple dies:).

I still can't believe Meni, the Man of Lightwave, is planning on jumping to OS X soon. I didn't realize it, but he was working at Station X when Fori and them were designing project:Messiah, as he had a large part in helping them design its functionality. His interest in OS X is very telling. Good days are ahead, my friend. As if you didn't know:) (Thanks Anon!).

27th April 2001

Anonymous Report on NAB: Lightwave

I received this fascinating anonymous report on NAB:

 

" First, it was very cool meeting Brad Peebler and Meni. Meni has got to be the Tiger Woods of Lightwave -- very easy to talk to and relate to while holding you in awe because of his superb skills. They both demonstrated (on the side) the two biggest upgrades to the 6.5 b patch -- one being the noise smoothing function which smoothes out noise from both radiosity and area light renders (and we're not talking a blur filter here, either, as everything stays sharp).

The other is the improved operation of bones, where the falloff can be set to the 64th power now. Meni is working on an animated short (The Creep, which he'll be posting the trailer for on his website once he gets the time), and all the animation of his little character was being done without the use of weight maps. There was still some subtle bleeding over of mesh influence from the bones on things like fingers, but he expected it to be completely eliminated by the time the patch was released. As a result of the reduced influence of the bones, the IK should calculate a lot faster as well!

He also talked about this patch being optimised for the P III and P IV, but I didn't catch the exact number. There's a link from Flay's website to Safe Harbor where they're claiming a 100% jump in performance with the P IV, and if they're right then it's time I started freshening up on sign language!

I will tell you this -- I get to compare a dual 500 G4 against an Athlon 900 at work, and I've found that for radiosity or area light intense scenes, the duall 500 is 1.3 times faster, even though its combined chip speed is only 1.1 times greater.

Tonight, while setting up another render, the dual 500 was rendering in 17 seconds what the 900 MHz Athlon was taking 44 seconds to render (which shocked me, making me double check the settings, and do an image to image comparison to make sure I hadn't missed something). So I think the G4 can hold its own, assuming you ignore cost of machine of course :)

Back to Meni's on-the-side demos, he was showing a scene from his upcoming short of an overturned garbage can on a concrete surface in a grungy part of town that he'd rendered with radiosity and noise smoothing.

However, the radiosity-rendered scene had been baked onto the objects in the scene so that you could view the scene with photoreal textures in opengl and get incredibly fast render times from it. When I first saw the renders, they looked like real photographs mapped onto his objects. By selecting what aspects of a surface got baked, he was able to bake in the lighting characterstics he wanted, while letting shadows, reflections, and specularity still respond to a changing environment.

He and Brad both commented on how the baking of surfaces, originally intended for gamers, has become a very valuable tool for creating realistic environments that render very quickly. They also talked about the next patch being made available in about three more weeks.

I don't know if that should be repeated, but I think just fixing the bones is worth that wait. Meni said that Alan Hastings had recently informed Brad that he had found an anomoly in his bones formula which he is in the process of correcting -- possibly eliminating all unwanted bone influence for good. Then weight maps can be used as organizational tools instead of being used to prevent unwanted influence in an object's mesh on a bone by bone basis!

This is going to be a good year, indeed! Just thought you'd like to hear this news." (Thanks Anon.)

Links

Newtek's NAB Site