Saturday, April 16, 2011

Blender 2.57 Released! Fourth Generation of BtoR is near!

At long last the first official stable Blender has been released since the refactoring began over two years ago. Many changes have been done to get to where Blender is now and of course more is yet to come, however this marks a milestone in the development of the next generation Blender to Renderman tool set. RIBMosaic has been worked on some more, with Jeff Doyle picking up the project and so far doing some fantastic work in getting the code to work with the new API, since the last version released by Eric was built before the recent API modification. I have not tried it myself yet but I would assume that as of right now a functional modern pipeline with Blender, RIBMosaic and Aqsis has been laid down, now it's just working on beefing up the features and the long road of bug fixes. Do not quote me on that though.

So this means that Blender to Renderman has evolved. This is not the first such occurrence though, in fact it would be in it's fourth generation now, at least since the turn of the century when Aqsis and Pixie were first released to the open source world. The first such generation was the Blenderman script, which functioned at a basic level but was not so useful for animation and was bug prone, however it did find use in film visual effects over in Spain, which so far has been the only known such instance of professional work using these tools. The next generation was the Neqsus script, which I still have a copy of and was so unique in it's design that it deserves to be respected to this day. It is the only known Blender to Renderman plugin that allowed Renderman previews of shaders to be displayed inside Blender itself. Neqsus had it's own GUI system seperate of Blender, which is how this was able to be accomplished. As with the previous generation this tool did not allow for animated exports, only still frames and after Blender 2.44 came out the plugin failed to work anymore. Bobby Parker, who wrote the Neqsus script, stopped working on it and pursued ventures elsewhere but remains active in the BtoR community. Of course the third generation was the first version of RIBMosaic, a huge milestone in the history of BtoR as it offered the most complete solution to get Blender data to Renderman format. This tool changed the playing field of the community, now we had a valid production worthy tool that was everything we could want, well almost. To achieve the level of intergration of Houdini or Renderman Studio for Maya was not possible, due to the restrictive PythonAPI at the time. It did however do everything else we wanted it to, we owe our entire existance on the web to this tool. I think without this tool Blender to Renderman would have died out years ago.

So now this is the fourth generation, Blender and RIBMosaic as well as Aqsis. Aqsis has become the strongest ally in this community, they are a devoted bunch of developers that have been very supportive of the whole Blender to Renderman idea, in fact the whole reason Project Widow started was because of them. Pixie has kind become stagnant in the area, where Aqsis has taken the role of leading the pack. While the feature set of Aqsis is not as complete as Pixie, 3Delight or even Pixar's Renderman, the developers have been working hard at getting there. On the other hand Aqsis has things going for it that none other can touch, for instance the Interactive Viewer of the prototype Aqsis 2.0 core, something not seen by any other RiSpec renderer so far. GPU relighting tools exist yes but the video demos done last year show just how fast Aqsis will be running on the CPU, which of course is where rendering for film takes place. GPU relighting tools emulate RSL code with GLSL code, they also use GPU rendering methods which are not the same as CPU rendering methods. Things like this make GPU relighting tools very handy but the Aqsis demos were rendered on the CPU in near real time, something the original engineers at Pixar had imagined possible long ago but knew it would be decades before it happened. Strange twist of fate though when it was an open source program that achieved this first rather than the inventors of Renderman.

Project Widow opened the doors to other areas of Blender to Renderman that had not been in the original equation. Areas like compositing, project management, collaboration tools, software development and pipeline design. In the beginning of BtoR it was all about how to translate Blender data to Renderman data. Step by step this was figured out. Then it was how to animate it, that too was figured out in the third generation. Technical direction found the need to get AOV in OpenEXR, only multilayered so that it could be imported back into Blender for post processing in the same manner that it does with it's own rendered frames. This lead to the development of shader tools that would generate complex shaders that literally would plug into the pipeline without problem, which in turn would lead to the capability of AOV rendering and compositing. Project Widow really opened up the need to see the pipeline as a whole rather than a section of it, simply because a lot of it really depends on each other. These are the things the animation and visual effects studios don't tell you, the little details like that which make production easier to maintain, instead they just say "We have these set of tools that are custom built to make our jobs easier"... but they don't go into detail as to WHY, much less HOW. These are known as "trade secrets". Project Widow is not professional by any means but the production is an open one, in fact tools and tutorials on how to do everything else BUT export Blender to Renderman data has been written. As mentioned before that was figured out long ago.

So now in the fourth generation, everything has been rewritten from the ground up; Blender, RIBMosaic and soon Aqsis. This website will also undergo that process. As things change so do we and plans are in the works to change this site again. When this will happen is uncertain, though in some areas it already has begun. Facebook has replaced Myspace in the "social networking" area of, Myspace seems to be lacking in areas such as point of existance but whatever Facebook certainly has proven to be worth the effort. The Project Widow Twitter account has been slightly useful, sending out impromptu screenshots or WIP renders for the sake of doing it with no regard to the fact if anyone would be interested, however it seems to be doing a decent job since it attracts more and more followers. There is no cohesion though between areas, such as the outdated and incomplete wiki, the groups... so on and so forth. Project Widow has taken up a considerable amount of time to complete which means in other areas, such as this site, things are not as complete as I want them. It happens. I just keep looking at the stats though, we are nearing the 100,000 visit and that is impressive for something that started as a small idea in 2005 after one ambitious young man had a movie idea and asked me to try to use Blender as the vfx tool of choice. Now several animation studios have made visits, motor companies, computer companies and countless universities worldwide. We have passed the point of crawling but we are only just beginning to stand. There is much more to do.

Friday, April 01, 2011

ILM open sources GPU based relighting tool

An industry insider has reported to certain people that ILM is putting a GPU based re-lighting tool into the open source world, a tool that has been used on many films by ILM technical directors and has been kept a secret for the better half of the decade. However ILM has already built a new engine that is based on REYES and GPU re-lighting code that had been part open source and part science research, so this leads to wonder if they are borrowing existing open source code and heavily modifying it.

This old engine is based off of technology that existed around the turn of the century and while it still functions very well in modern pipelines, the code is old and ported from old SGI libraries. The new tools currently in development are built from modern libraries for Linux, thus removing the chains of legacy obsolete code. ILM is presenting this code to the world mainly for the hell of it, they no longer feel that there are any secrets worth keeping. While the code may be functional, the fact remains is that it is obsolete and much of the pipeline is lightyears ahead of what had been created even a decade ago.

More information as to the location of this tool will be revealed soon.