One of the tough parts of working in 3d is knowing when to call something “done”. Every project has room for improvement, and it can be pretty tough to know when you’ve hit the point of diminishing returns.
I’ve gone through several dozen (perhaps as many as a hundred) iterations of the the rust texture, trying to address problems I’ve perceived with it being too clumpy or not spotty enough. It was also pointed out to me that the rust might be more realistic if it retained some specular highlights.
I found several ways of addressing both of those problems, but in the end, I liked the rust with very low (almost nonexistent) specularity and the large patches you see in the video.[vimeo http://vimeo.com/35887864]
Maya was crashing when I tried to render on the school workstations. After way too much troubleshooting, I happened on this blog at Autodesk, indicating that Maya 2012 will crash when rendering with Mental Ray when encountering a luminance node linked to a bump node.
As you can see from the image, I had an offending luminance node in the bump for the rust part of the layered rust shader.
The good part of this is that the troubleshooting led me to clean up and rename parts of my shader network.
I really appreciate Autodesk’s online documentation of this bug – without that, it would have taken me significantly more time to isolate.
Since I budgeted extra time for this first milestone, I thought I would revisit the modeling and fix some problems that were present in the version of the helmet I did for ANI 155. These problems were details, but for me, they were causing problems with realism.
The problems I wanted to address were:
- the shoulder rim edges were perfectly sharp
- the helmet met the shoulder girdle with no weld, roll, or seam
- there were some floating screwheads from the previous helmet stand that needed repositioning
Additionally, there were still parts of the rusty metal texture that I wasn’t all that happy with – for one, the bump seemed to eat into the surface of the helmet too much. Previous versions had used separate shaders for the helmet attachments/protrusions, which I wanted to consolidate. On the technical side, I simplified the bump network with Layered Texture nodes instead of Layered Shader nodes.
Last, I wanted to create a new bump map for the wood texture that really drove home the sense of age that I was looking for.
I’ve also been experimenting with render layers and render passes; I’m still working out exactly which passes that I want – I’m leaning toward Ambient, Ambient Occlusion, Diffuse, Shadow, and Specular, although that might change based on usefulness in my experiments.
Additionally, I’m experiencing some definite problems with going back and forth between Maya 2011 and Maya 2012. Right now, scene files that are rendering with no errors in 2011 are causing Maya 2012 to shut down.
I hope to upgrade my home version to 2012 after this project; since that will bring me version-current relative to the labs at school, hopefully that will cut down on errors and increase my efficiency in completing these projects.
My goals for Milestone 1:
- have diving helmet rendering with sun and sky
- improve/rework workbench texture (looks low-res on the stand)
- start setting up breakdown shots/render passes (diffuse, spec, shadow, AO, etc)
How it went:
Sun & sky turned out not to be the best lighting solution here – there just wasn’t enough environmental detail to make the metal reflections look good, and putting the light outside the window killed the nice specular highlight that the area light was giving me. So after returning to the previous light set up, I tweaked the FG settings and ditched the depth of field I had been trying out. This cut my render time down to under 3 minutes per frame for the master beauty pass.
When I rendered out an image sequence to test how my lights held up, I noticed two problems: first, the shadow samples were too low, giving the helmet’s main shadowed area visible moving noise. I remedied this by increasing the shadow samples of the main area light from 8 to 16. This caused a modest increase in rendering time, but kept the beauty pass under 4 minutes.
The second problem that the image sequence revealed was, of course, the normal FG flicker. Since the camera was the only moving object in the scene, this was remedied by freezing the FG maps and tweaking the FG settings slightly.
In addition to the lighting adjustments, I added some color variation to the rust texture (circled in the old render) and adjusted the bump map to look more natural. I also remapped the wall, workbench, and helmet stand to address the slight low-res distortion that I have circled in the old render image.