Renderman “Real-Time” Compositing (Results)

Initial Tests:

Fortunately, I was able to setup the Interaction Tools on the school’s machines with little trouble. I then began working with the examples provided by Hal Bertram to piece together how to use his system.

First of all, I was able to run composite tests with ambient, diffuse, specular, and light passes as well as light attributes with very few problems. However, it took me a few tries to realize that it wasn’t re-rendering the TIFF files each time. In order to get it to generate new TIFFs from the RIB files, and therefore change the renders, I have to trash the prepare.tif file each time.

I’m planning on changing the MakeFile to delete the prepare.tif file within its folder each time I run the program.

Here are some of my implementations of Interaction Tools, using a few of my models from classes:

I have a few quoms with this method of compositing. First of all it’s still flat. Essentially, I’m compositing stills together, which doesn’t give me the freedom I was hoping for with the rendering process. Secondly, the quality of the renders are not where I wanted them. I plan to work more with my render settings and see if there is a shading rate that is responsible for the low-quality renders, or if it’s something to do with the halfway conversion to CG.

Surface Attribute Meshes:

In his extension of the Interaction Tools to deal with geometry, Hal Bertram devised a method of baking point and color data (surface attributes) to file, in what is essentially a surface attribute stream. (SAS) Included with the Toolset is an executable file that converts this stream to a structured “mesh” file, or surface attribute mesh. (SAM)

I knew that as of the alpha release of Interaction Tools, SAMs were not compatible with PRMan 13.0, but I was hopeful that 13.5 might be more forgiving. Unfortunately, the conversion executable throws a St9bad_alloc error whenever I run the MakeFile. I hope to get more information this week as to how I can get this up and running, but for now I’m at a standstill.

Until I can get Surface Attribute Meshes to work, I will be busy converting some of my prior RIB files to work in Interaction tools. In this way, I hope to improve the lighting placement and compositions to further polish my renders.

Conclusion:

I have made contact with Hal Bertram in order to get some input for these solutions directly from the source. Unfortunately, as of the end of the quarter, our conversation has not gotten past that initial contact.

The research I was able complete in the three weeks alloted was enough to give me an idea of what such a pipeline would entail. I do think that given the right circumstances, a rendering pipeline could be streamlined by the implementation of Interaction Tools. However, I’m not satisfied by the overall low quality of the renders from the system. Ideally, I would like to have intermediate to good quality for the real-time composites. Otherwise, it would be difficult to predict the final renders..

Updated Task List:

1) Continue research into implementing SAMs
2) Modify prior RIBs to further polish renders
3) Generate a Geometry-Based Composite for tweaking camera position

Possible Additional Goals:

4) Generate occlusion in the pipeline to tweak samples in real-time
5) Test volumetrics in the pipeline for light fog