|Page (1) of 1 - 08/02/05||email article||print page|
Face Time with SOFTIMAGE XSI v. 5.0Preview of an impressive new facial animation system
On the eve of SIGGRAPH 2005, I sat with Softimage senior product manager Gareth Morgan and product specialist Erik Goulet for a look at some of the impressive new technology being introduced in SOFTIMAGE XSI 5.
They showed XSIs new 64-bit ready gigapolygon core, which can handle up to a 10x increase in detail, along with a new memory management system that makes it possible to render any scene that can be loaded, regardless of physical memory limits -- even on a 32-bit laptop with 1GB of memory. They showed Gator (Generalized Attribute Transfer Operator), which transfers surface properties of one model to another a dancing clowns behavior can be transferred to a lamppost. And they showed new migration tools for Maya users that provide familiar navigation for those whove got Maya menu, keyboard and mouse patterns hard-wired into their brains
But one of the most intriguing demonstrations was of a new facial animation and rigging system being shown at SIGGRAPH as a technology preview, with plans for a fall introduction. Softimage technical staff have been working with L.A.-based Blur Studio to develop a detailed computer model of facial soft tissue. This led to the Face Robot (aka Rock Falcon), whos got a subtle range of built-in expressions and facial deformations. Character animators will be able to use this technology to create high quality facial animations much more quickly than using traditional facial animation methods.
|Face Robot controls.|
Face Robot, which supports both keyframe animation and motion capture, was on display at the SOFTIMAGE booth, and Blur Studio created a short using the new facial animation technology.
?In the Tom Hanks movie, Polar Express, they had to use several hundred markers for the facial capture. The technology that weve developed for the facial animation system actually introduces algorithmic procedural soft tissue models, to do a lot of that facial tissue movement. So you need, say, 20 or 30 markers for a face, to get the same or better results, Morgan said.
?This means you dont have to have the voice talent in makeup for four hours getting all the markers put on, you can do it in 10 minutes. You can have them for four hours more doing dialog. And thats a key part of where we see character animation going, Morgan added.
He noted that whats driving such innovation is Softimages philosophy that the software should solve the complex technical problems and open the door to creative production using the tools let artists be artists.
|Range of Face Robot expressions.|
That could result in higher quality animation, too.
?If youve got that much more time then can refine it even more, Morgan said. ?Its not in a day or two that youre going to get Gollum, its not the same thing. But youre going to get significantly further along the way. So you can take the skills and talents you do have and use that to iterate and refine.?
At the conclusion of the demo, product specialist Erik Goulet showed off the new memory management system, opening a detailed model of an elephant on a single CPU 1.7 GHz laptop with 1GB of RAM, an ordinary 32-bit machine, and began rendering it at increasingly higher resolutions. He first subdivided the elephant with a geometry approximation (sub-D) and it went to 37,000 polygons, or triangles. Then he continued to subdivide it higher and higher.
?If we decide to go into the higher end, well have tons of polygons, and this elephant will become completely white, because there are so many subdivisions to it, Goulet said. So what Im going to do is switch it to the bounding box, and were just going to keep on cranking this and Im going to go all the way up to 5, where were going to have approximately 9 million triangles. So for a little laptop, thats not bad.
Then he set it to 7, subdividing the elephant into 152 million polygons. Before drawing the Render region, he went to Options and set it to Heavy Scene Model Optimization for Scalability. Then, with the Quick Set option selected, mental ray set itself in the appropriate mode.
?So when youre going to do a Render Region like this, regardless of the amount of memory that youve got, its going to go tile by tile, using the memory, and when it goes to its cap, its going to flush all of the calculations that have been done previously that it doesnt have any use for anymore, Goulet said. ?So its going to regulate itself. Thats why we say when you can load a scene in XSI, you can render it.
As we left, the laptop was steadily chugging away on the 152 million polygon render.
Related Keywords:Softimage, SOFTIMAGE XSI 5, facial animation, rigging, SIGGRAPH, Face Robot