Cloud System Prototype
Generating realistic looking clouds and lighting them properly is an exceptionally complex task. The appearance of a cloud in the real world is a result of the interaction of an incomprehensibly large number of photons with an incomprehensibly large (not as large as the number of photons, but still incomprehensible) number of water/dust globules. No computer in existence could accurately simulate a weather system with that level of detail and even approach real time rendering speeds (you think 2 frames per second is bad, try 2 frames an hour.)
And so, the task of the real time computer graphics programmer becomes: find a way to simulate a system that looks like that hugely complex system but only requires the simulation of a bare minimum of the variables present in the original system.
There’s a system by WindwardMark Interactive called Nimble that does a beautiful job, but unfortunately, last year it was bought by Linden Lab (owner of Second Life.) Although the WindwardMark website doesn’t explicitly say, I imagine that means it will not be on the market for indie developer peons such as myself.
Once my research on existing cloud simulation systems was complete (and no suitable existing systems were available,) it was time to write a cloud system. I spent a lot of time thinking about cloud systems, but for a good while I couldn’t put pen to paper (metaphorically speaking.) After originating and discarding a handful of unique ideas for the cloud system, I decided to just start writing the simplest solution and see what it looked like.
It looked a lot better than I expected. It is far from done and there are a lot of things “wrong” with the clouds in these screenshots but it surprised me that such a simple approach could look even this good. Though, admittedly the clouds look a bit more like thick rolling fog than clouds at the moment, but that shall be remedied soon enough.
Tags: 3D Graphics, Cloud System, Dynamic Clouds, Dynamic Sky, environment
May 2nd, 2008 at 12:40 am
I remember reading about Linden Lab acquiring Windward Mark. I’d just found out about their amazing technology; hearing that such a huge money-making corporation bought their product before anyone else could even try to use it, I could have cried. I think I feel some tears welling up right now, in fact.
But yes, I’m bored of skyboxes. Done nicely, they are efficient and functional. Even fading between certain skydomes for day and night would be sufficient. But dynamic clouds sounds so much more open and interesting.
I spend a lot of time studying the skies seen daily around the valley. There are a lot of things to notice, which commonplace perception will unwittingly overlook. It’s true that your prototype looks little like the real thing, but it is a great start, and one can tell what they are supposed to be: clouds forming in the sky.
What have you tried so far to light them? My knowledge of the programming for procedurally generated art is essentially non-existant, but perhaps to integrate image maps or simple, deformable spheres could help determine what is lit and what is shadowed. When drawing a form such as a cloud, the artist looks at its simple forms: an oblong sphere, perhaps, with most of the shadows here, highlights there. The random detail is then added to the overall shape, broken down into other components of form as necessary.
Though I really have no idea to what creative complexity you are interested in pushing the clouds. As a budgeted development studio would possibly say, they are “just clouds; who is going to look at them, anyway?” I am, but then, my opinions on what’s interesting and beautiful are hardly normal.
June 5th, 2008 at 2:33 am
[...] cloud system has come a long way since I first posted about it. I suppose technically, it has come a long way since yesterday when I started working on it again, [...]