What have I been doing for the last month? Rigging, scripting, and blendshapes. Rigging our main character, Arthur Winters, has been a right pain to do. It actually started off quite fun when things were just working, but then after about 2 weeks of rigging and dissertation, and almost nearing completion, things started to go wrong. And so for the last few weeks it has been a long tiresome battle with blend shapes and scripting. Not that I don't like scripting, its actually quite enjoyable when it works!
Anyway, I will be showing the rig in great detail when it is totally finished (which I hope will be soon, considering that we all want to animate by Thursday).
I do however have something else to show today. While rigging, I have done a few other things toward the project. Firstly, some lighting tests, and secondly some liquid dynamics. Jordan has been busy modelling and texturing the main set, and in fact you can see the render he did using Mental Ray's Final Gather lighting on his blog. Visually it looks great, and so the only real problem with this was the render time. We don't have the time to render at 4 and a half minutes a frame (which is how long that render took), especially considering that the render times will increase with the rest of the props and characters are added, so my goal was to use this image as a guide to faking this type of lighting. During the summer I came across a few new lighting techniques which I'm now using everywhere! I'm not going to say that these are our final lighting setups just yet, but here are some of the results of the tests.
Day
Night - (there is an imaginary lamp light by the window)
Possible dream sequence lighting
and slightly darker... possibly too dark actually.
The exterior (no textures yet)
Apart from the exterior, (which took slightly longer) they rendered in about 50 seconds each.
Onto liquid dynamics. Now I've actually been familiar with Maya's liquid nDynamics for quite a while, and have probably done what seems like a million tests already before this project, and realise what it's limitations are when compared to other (more expensive) programs like Real Flow. The first major limitation I've found is its speed when calculating with millions of particles. The main issue is that nParticles aren't multi-threaded, meaning that it can't use multiple processing cores that you find in computers these days. The simple answer to this is to sacrifice the quality by using less particles.
The next issue I've found is the mesh that Maya generates around the particles for rendering a solid fluid. No matter how much tweaking you do, it always looks lumpy, like porridge. I don't think lumpy whiskey is going to look very good, so it was imperative that I find a fix for this issue, and I don't think I would have found the answer unless liquid dynamics was needed in our animation.
I very useful node called the peak deformer from the free plugin SOuP provided half the answer. It makes the blobby appearance of the liquid sharper, by creating peaks. The other thing which improved the mesh was to use Maya's own average vertices tool. This helped to smooth out some of the blobs and the peaks to create a liquid shape which looks a little more like water or whiskey.
Before
After
The video
It's never going to look 100% perfect though, but its a darn sight better than this...
Of course these were only tests, so I'm going to have to recreate the sim when we have the scene finished and the bottle correctly animated. Apart from that, all it needs is a good shader!
Now I think it's about time I get on with finishing the rig… once and for all!
A Flock of Pixels
whoop whoop! :)
ReplyDelete