Clouds theory. [Some good solutions how to creat clouds]

One of the first problems that had first games is how to create clouds. Run-time or pre-generated, it’s still very difficult.


The first good solution were sky-boxes, widly used even today. But come on! It’s 2009, not 1999 :unamused:… The next solution isn’t used at all, but I like it. It’s been made by Mark Harris and was revolutional at that time. It’s called skyworks and is dynamic(well not that much) & realtime. But it’s pre-generation alghoritms are really slow, and 36 clouds take already too much time. And implenting it to Panda is, I think, too hard work for most people.


Mass cloud pre-render(Thomas’s method):

Pre-render thousands random clouds and make the randomly spawn everywhere. Deform them if the main object/camera moves, or apply dynamic bump maps on them(may be simmiliar to water dynamic bump-effects). Then apply the billboard effect.

Dynamic sky-boxes[[color=red]!!!WARNING[color=red]!!! CPU INTENSIVE](Richard&Francis’s method):
Make some cloud-layers. Apply them on the skybox, so that it’d feel like they would be volumetric.

Volumetric clouds[[color=red]!!!WARNING[color=red]!!!] GPU INTENSIVE](Crytek’s method):

(in buffer)Make an ellipse. Extured it in all sides by pi. Join all vertexes and create some new quads from them. Repeat this until you get a full polygon with 2 layers. Delete everything uneccessary. Take the cloud-texture, desturate&blend it to the limits and put it on the model. Make the model fully transparent and put the half desaturated texture on it. Make the cloud texture transparent in post-process. Rotate the model and take the pictures to the memory. When the function is called put the needed picture to the screen.

Mega craziness[Warning! For experimental uses only!](Dany’s method):

(in buffer)Create a black circle. Add lots of other small circles, but semi-transparent. Create a random white, desaturated, and transparent polygon and put it on the created picture. Repeat it as many times you want.[size=59] Here are some facts: if you make a small picture, and repeat it much times it gets more realistic, but less nice. if you make an enormous picture and add another layer, it gets even, even more realistic, but the less nicer. BUT if you make lots of medium-big pictures and the count of repeats would be dynamic, you get realistic, nice, and fast![/size] BUT you must create a cloud-limits polygon, but so, that it’d be nice, and this is pretty hard! But after you would make it, it’s fun :slight_smile:

I hope it was it was helpful. :smiley:

The solutions have nicknames, there’s no Thomas, Richar&Francis. Don’t get confused!

i just love to post this link over and over again for everyone who haven’t seen it yet.
features most of the currently known algorithms and implementations.


Why I’ve never seen this helpful resource?

btw I like this one: