This article focuses on shading and texturing the Mini Cooper, I’ll talk a little about rendering with Mental Ray but I think other people have covered that in more depth elsewhere.
I’d wanted to build a cg car for a long time, but couldn’t decide what model of car it would be. It was only after seeing a few New Mini models I thought I’d have a crack at the original model. This project would also be my first time using Maya so I didn’t worry about taking my time.
The car was modeled in the conventional polygon manner; every element was constructed from a single plane using the Connect Poly Shape tools written by Dirk Bialluch. A great tutorial for this method of modeling is available from Munkmotion: http://www.munkmotion.com/tutorials/nsx/
When I started work on the car I began also experimenting with the Mental Ray add-on for Maya. To be honest it was pretty clunky and extremely slow. The documentation was useless and there wasn’t much in the way of tutorials of help on the web. I tinkered around with it and tried to understand the Global Illumination and Final Gathering systems.
I found the early implementation of Mental Ray for Maya extremely frustrating and had almost given up when Alias got their act together and brought out the new cleverly integrated free version. Although still tricky to master this version was a lot faster.
I decided fairly early on that the car should be muddy and dirty. This mud would need to be textured onto the car from bitmaps, which would mean huge file sizes if I wanted to render close-ups. I needed a system to allow me to tile and reuse my mud textures so they could stand up to extreme close ups.
Painting textures into simple Photoshop files and just applying them to the car wasn’t going to work.
The solution I settled on was a fairly simple one, using masks to define where the mud would lie on the car, and then mixing through to a mud shader in those areas. The mud shader was mapped separately and could be tiled and varied without affecting the paintwork.
I separated the dirt into two layers a dark wet mud and a thinner dry dust spray, each layer had its own mask for each body panel. The mud itself was mapped globally to the whole car with a tri-planer projection map. This way the mud would run from one panel to another neatly.
Mostly importantly this approach meant that I could deal with the look of the mud separately from its placement.
Once I’d UV mapped each body panel I was really to paint my dirt masks. I used Maya’s 3D paint system to paint myself guides on each panel, and then exported these TGA’s out to Photoshop and painted more finished versions.
I also exported the UV coordinates at the same scale and put them into the Photoshop file, this gave me exact reference of where my paint strokes would appear on the car in Maya.
I then worked on two masks for each panel, a basic dry dirt mask and a wet splatter mask:
The black areas define where the mud will sit and the white where the paint will be.
All the masks were hand painted and assembled from various sources. I used a lot of concrete texture maps layered up and manipulated to create the basic dirt masks.
The red paint shader was fairly simple, although given the strange way mental ray’s FG system seems to deal with shaders it took a lot of tweaking to look right. All the lighting in the scene came from a HDRI map.
The shader is a simple red Blinn with the diffuse component turned right down to 0.4. The reflectivity and specular colour have a sampler info node passing through a curve to really push the specularity when the metal bends away from camera.
The curve looks like this :
It’s essentially turning the linear 0 -> 1 output of the sampler info’s facing ratio into a 0 -> 2 exponential value.
I use curves instead of ramps to do this as I’m used 3D Studio Max’s way of doing things, but really it makes no difference how it’s done.
I’ve had a lot of emails asking me how on earth I use curves in the hypershade, so I’ll attempt to explain here.
1 – Go to frame 0. Create a blinn material, open up the attributes and right click on the specular colour attribute. Select ‘Set Key’ from the options.
2 – Now press show upstream and downstream connections and you should see three animation curves powering the specular colour. Delete two of the graphs.
3 – You should have the same as I do above. Drag and connect the curve to both remaining colour inputs on the specular colour.
4 – With the curve node selected open up the Graph Editor. You should see just one single keyframe sitting at 0. If it’s not at zero horizontally (usually time), place it there. Create another keyframe at 1 and set the value as 0.25.
5 – Now that all remains is the plug in the facing ratio from a sampler info node. Create a sampler and drag to the animation curve, select ‘other’ and connect the Facing Ration to the Input of the curve.
6 – Done !, you can play with the the values on the curve and check the effects.
Because all my muddy areas were defined by masks I could tweak the actual look of the mud over the whole car without having to keep going back to Photoshop. I mapped the two types of mud over the entire car as a Tri-planar projection, and then experimented with adding gradients to have the mud at the bottom of the car darker (wetter) than at the top.
Many of the settings on my shaders were connected to a control object in the scene, by changing those settings I could alter the look of the paint and mud instantly. These global controls really sped up the process at the end, even if they made duplicating shading networks a pain.
An early experiment using a different type of mud. This looked like someone had chucked a bucket of wet concrete on the bonnet.
As you’d expect rendering took a long long time. My frames were averaging one hour on a 3Ghz machine, and I ended up rendering an animated sequence of about 1000 frames !. I was lucky to get the use of the machines in work.
Mental ray had some serious flickering and crawling issues. Setting the final gathering min and max radius correctly really helped prevent this. According to the mental ray information I found the max radius should be set to 10% of the scene width, and the min radius should be 10% of that. I ended up with 20max and 2min
I also played with the BSP settings to speed up render time, I’ve no idea why but I had most luck with BSP size set to 16, BSP depth set to 60 and BSP max memory set to 0.
Some renders still had a little flickering so I had to render out small fix up patches to place over those areas.
So there you have it, a little bit of behind the scenes info on the shading and rendering of my Mini Cooper.
Click here to go to the gallery to see more cooper images.
I’m a highly experienced freelance CG & VFX Artist. In the Commercials world I’ve led many award winning projects; both as an onset VFX Supervisor and as a CG Supervisor. I’m now branching out into Film VFX and am currently at Double Negative as a 3D Generalist. read more