studioGPU

Hardware tesselation brings the artist vision into alignment with the final product

SIGGRAPH 2009 was an exciting show for StudioGPU.  We showcased the upcoming version of MachStudio Pro V1.2 and announced that MachStudio Pro now ships with the FirePro V8750 2GB accelerator from AMD.

But the fun part of the show was watching peoples’ jaws drop when they saw MachStudio Pro in action.  Everyone understands the concept of a really fast final renderer.  But real-time workflow, working creatively at final render quality was what drove the “I don’t believe it” reactions.  The best way to show you what I mean is to let you watch a demo we captured on camera.

The demo runs through real-time ambient occlusion,  lighting, on-the-fly shader adjustments and finally shows sub-pixel displacement mapping using hardware tessellation. 

Tessellation involves breaking down polygon meshes into higher poly meshes. In the video below, taking a simple polygon mesh and breaking each polygon down into smaller units allows the surface of the object/model to become smoother with a greater sense of depth and detail, including self-occlusion, self-shadowing and silhouettes.

But more detailed characters, especially with complex animation, devour memory and up storage requirements. Software runs into bandwidth issues so load times increase and memory demands shoot up. But the new version of MachStudio Pro takes advantage of the FirePro V8750 GPU to generate the additional complexity using hardware tessellation.

Hardware tessellation means there is no penalty hit for the virtually unlimited fineness and detail created by the displacement maps.

Quick note:  Once you start the video, click on the HD button on the lower right, to watch the video at higher quality.

So just who are some of the people behind StudioGPU?

MachStudio Pro was the brainchild of a group of digital artists and engineers who had worked together in the game industry for 15 years. So who are some of these folks?

One of them is Chief Scientist Yoni Koenig.  The crew at AMDUnprocessed came into the Hollywood office and did an unrehearsed interview with Yoni.  It is an interesting watch and listen as Yoni talks shading pipelines, lighting pipelines, compositing, render farms, Hollywood and more.

There are also interviews with David Koenig - Chief Executive Officer and Andrew Baum - Vice President of Marketing.

Shaking up the time vs. quality tradeoff in the stereo 3D production pipeline

Unless you've been living in a cave the last few years, you know that Stereoscopic 3D is one of the hot topics in CG animation for cinema. Think Up, Ice Age 3, Journey to the Center of the Earth, Toy Story 3, Monsters vs. Aliens, G-Force 3D, to name a few (see more). The market for 3D stereo gaming is probably even larger, with both the X-Box and PlayStation 3 expected to roll out upgrades this year to support stereo 3D. If the topic interests you and you are attending this years Siggraph, be sure to check out the Computer Animation Festival which is devoting an entire segment to stereo 3D.

I've been following stereo 3D for years, but primarily for scientific visualization, where artistic quality control and live action footage were not issues. Then a few weeks back, I came across a twitter reference (@sterlingy) that piqued my interest. The tweet was by the guy (Sterling Youngman) who is behind the Pansy Warrior Princess (PWP) - a somewhat unusual and definitely niche site, that has a pretty sizable fan following, having surpassed one million viewers (reminds me of a cross between Monty Python, Elvira, Mistress of the Dark and the recent World Of Warcraft Mountain Dew Game Fuel commercial).

Youngman is producing a Pansy Warrior Princess 3D stereoscopic music video to debut at IBC this September and is using MachStudio Pro as the lighting and rendering solution to keep both time and costs under control without sacrificing creative control.

Stereo 3D production without the big studio budget

The entire PWP music video project will be CGI, except for Pansy and her band, (which will be shot on green screen using Silicon Imaging’s new 3D 2K camera system). The original plan called for rendering the CG with Brazil. However once Youngman committed to going stereoscopic, it was clear that lighting and rendering the CG under the time constraint of being ready for IBC, would not be feasible without a big studio budget for lighters, compositors and render farms. Enter MachStudio Pro.

Details about the technical and 3D production complexities and workflow will likely end up in upcoming magazine articles after the video debuts. So for this post, I am just going to touch on some high level points.

Anaglyph from a preliminary scene rendering. If you have a pair of red-blue 3D glasses, check out this 3D CG image. In the production, MachStudio Pro is used for lighting and materials both in pre-viz and in final renders passes, with render times reduced from several minutes, to a few seconds .

Pre-Viz

Youngman is a 3ds max guy. The scenes he created were moderately heavy. So in his words, doing the pre-viz in Max, even without any significant materials or lights, just to determine the suggested camera moves was “not pleasurable”. Once he exported his scenes into MachStudio Pro however, he could apply lighting and materials all in real time, zipping around even with AO, and reflections turned on - all without having to wait for a lengthy render to view the final result.

Live-action to final composite

After filming the live-action footage of the video in stereo 3D, the pre-viz cameras are dumped, and the live action stereo cameras are re-created in Max . Those cameras are then exported back into MachStudio Pro for rendering left and right backgrounds. Light rigs and materials that were created in MachStudio Pro during pre-viz, can be further adjusted as needed artistically (and applied to whole scenes rather than shots). Then the CG is exported out in passes for final live action compositing and touch-up. Keep in mind that stereo 3D requires two renders to achieve parallax - one for the left eye and one for the right eye. So shaving final frame render times from several minutes down to a few seconds was integral to making this project possible. If all goes as planned, expect to see the PWP 3D video this September.

Enabling the small studio and creative artist

Beyond the production story, what I found really interesting about this project is that Youngman is producing a studio-quality 3D stereo video without the backing of a big studio or big investors. He caters to a niche audience and by combining a lot of determination with several pieces of new technology and integrally with MachStudio Pro, he is able to compete on a creative, production, and time level with much larger and better funded ventures. In many ways, this shakeup in the time/cost/quality tradeoffs are reminiscent of smaller studios using Final Cut Pro to compete with high-end editing systems in big post houses, or PhotoShop artists taking on tasks previously reserved for Quantel mavens. The technology becomes a great equalizer. Catering to niches becomes viable and profitable.

I look forward to watching the PWP 3D video, but even more, I look forward to seeing all of the other 3D CG animation (niche and broad reach) projects it inspires.

About the music

As a way of note and back story, the song that inspired this video was written by Deirdre Broderick with production and guitar solos by guitar icon and legend, David Spinozza. If you are a metal head or a Pansy fan, check out the MP3. Also if you are in the LA area, and interested in watching a the live-action filming, it is taking place next week (June 13-17) at Panavision's sound stage ( email for details).

Shot-based vs Scene-based pipelines for the lighter/compositor

If you are a lighter/compositor working with desktop rendering software, then you are used to working in a shot-based pipeline. Shots are set up, animated, lit, and rendered on a single camera move/shot basis. For example, you might light and render a long camera pan of a Cylon Raider traveling through space (I admit I was/am a BSG fan).

For a a typical animation scene (e.g. Cylon Raiders attacking Colonial Vipers against a star backdrop), the standard approach is to break the scene out into multiple shots, then distribute the shots to your lighters for setup and rendering or setup and transfer to a render farm.

A lot can go wrong

A lot can go wrong with a shot-based pipeline - from continuity problems where two different lighters work on shots that cut against each (each having their own "artistic" eye), to different lighters not understanding your shot breakout. In either case, you end up wasting render time twice over (once for the mistake and once to fix it) as well as wasting the time and goodwill of your lighting team.

Even if nothing goes wrong, a significant amount of time and energy gets absorbed when lighters have to go through the pains of recreating light rigs across shots or go through an import/export procedure to get those rigs into several independent scene files.

But because rendering is typically an action discreet from lighting/materials setup, this type of shot-based pipeline with multiple lighters/compositors is necessary.

The scene-based pipeline alternative

But what happens if the technology changes so that you can work in a scene-based pipeline? In other words, what happens when you can use MachStudio Pro, where the animation and lighting you see in the viewport is the same as the final render passes?

In that case the entire workflow changes. The same lighter can work on all of the shots in the scene using a common set of light rigs and render passes. This not only means increased consistency and reduced logistical errors, but it also means significantly reduced production time and production costs. No recreation of light rigs, no import/export of setups, no setup-preview-render, no render farms. Crunch time becomes a little less crunched.

For many scenes, you could even bypass compositing because the render pass could potentially include everything you might need including animated depth of field, HDR cameras, multiple light sources, AO, blooms etc. - everything that might go into making a final composite. Or when you need additional compositing work for special FX or live action compositing, export out the independent channels - but as an entire scene, not as single shots. This is a significant shift in the workflow and the artistic flow.

So next time you are doing a render, imagine what it would be like to work in scenes vs shots.