Process Blog for SDGM 560, a 10 week collaboration with The Mill at SCAD.
Week Six Updates:
This week has been full of troubleshooting. While we got more light in our last iteration, there was still no contrast and a complete loss of detail. While doing tests of adding more lights to see where we could pick up detail, we realized that the shaders weren't really responding to our lights.
We realized that our roughness maps were almost completely white, and were completely blurring all detail. To figure out what was going on, we began doing some tests with roughness.
Roughness set to zero (above).
Roughness set to zero, with a spotlight added (above).
Roughness set to zero with a spotlight, with addition of other assets (above).
The results were extremely strange, as the road and terrain were reacting very differently to the added light. We soon realized this was because the road had a very high roughness, and we were testing the terrain with no roughness.
While defining the shaders we had made the mistake of compensating for an IOR that was too high by increasing roughness. Therefore, we knew we had to go back and rework our shaders with a lower IOR and lower roughness.
We held off on re-adding in roughness, displacement, and normal maps and worked with numerical values to see the changes, and still were getting strange results even with low roughness and low IOR. The entire terrain looked very gray and almost frosted. We then went back to our quilt to see how the shaders were acting on simple geometry.
By looking at the specular highlight on the spheres, we realized what was going on. With a high roughness the terrain looked normal because we were blurring the highlights so much that only the base color was visible. Without any displacement or normal mapping, the terrain was essentially a flat plane and therefore the specular highlight was covering the entire piece of geometry, giving it a frosted look. We knew that balancing the normal mapping, displacement, and cat-clark subdivisons was going to be essential in maintaining surface detail for the light to catch on to add contrast.
Throughout this process, we also came to the conclusion that our chosen color palette and lighting situation was just not working. A backlit nightime setting was just not showing off our assets in the way that we wanted, and so we decided to switch our time of day and lighting set up completely.
We started by finding reference of the look we hoped to achieve:
We then searched HDRIHaven for an HDR with the right key to fill ratio we wanted:
HDRI by Greg Zaal from HDRIHaven.
We then implemented our new HDR, and continued reworking shaders.
First test with new HDR. (Note: the background sky is just the HDR out of Maya, and will be replaced with T'Naige's matte painting)
In order to optimize our use of subdivisions, we split the geometry into more pieces: foreground, middleground, and background. We gave the pieces closest to the camera the highest number of subdivisions and the background the least amount, while still ensuring that we were creating enough surface variation to provide detail in the terrain.
Renders with corrected roughness maps and added lighting (above and below).
After reworking Shot 1, I moved on to working on the new lighting for Shot 2 and Shot 3. I also implemented the updated shaders and continued to optimize subdivisions for each shot. I then passed my renders to Naige, so that he could integrate them with the road, effects, and foliage from Houdini by Elyse and Billy.
Updated full composite (above).
Old Shaders/Lighting vs Updated Shaders/Lighting (above).
This coming week we will be continuing to refine all of our assets as well as fix some timing issues with the effect and car. I will also be adding an additional render layer to shots 1 and 3 using a ramp shader, which Naige will be able to use as a mask in compositing to establish a sense of atmospheric perspective. Everyone has been working so hard and doing such an incredible job, and we can't wait to see this week's progress!
Hozzászólások