Quote:Original post by Specchum
That's a damn nice looking Sun. However, as a suggestion, perhaps you should cut down a bit on the dark bits and have a bit more of fire and orange stuff?

I think everyone is misunderstanding the image I posted above... this is a kind of effect I would like. Unfortunately my sun looks no were near that good. I was wondering if anyone knew how to do something like that.

Sorry for the misunderstanding.

Quote:Original post by Battagline
this is a kind of effect I would like. Unfortunately my sun looks no were near that good. I was wondering if anyone knew how to do something like that.

Looks like there are two main graphical effects in that image.

1. High Dynamic Range Imagery - Look at the bits where the white flare (?) is next to the spaceship, the glow it gets/overlaps the foreground slightly.

2. Procedural/Noise Textures - This I can't be sure of, but that "molten lava" like effect looks like it could be done quite nicely with a few layers of perlin noise (of different colours/frequencies) blended together.


Number-1 above can be done properly in realtime on the very high end hardware (those with good floating point texture support), but obviously that limits the number of people that can actually experience such an effect. It can be faked using standard 32bit textures and a few pixel shaders, or (as DrunkenHyena) mentioned, some clever tricks using imposters.

Basically, HDRI gets the glowing effect by splitting the LDR and HDR colours (e.g. takes the very bright sections of an image) and blurs them before adding them back to the image. All as a post processing effect. The net result is that the bright parts of an image appear to glow - and being in screenspace, you get the nice "overlapping" type effect I pointed out above.


Number-2 is a fairly straightforward algorithm, bit more tricky to get done in realtime but still perfectly possible. Procedural textures work quite nicely here - and, if you can crack realtime, then you get easy animation possibilities.. The number of samples required can be relatively small, as you can rely on the hardware to interpolate the results by filtering the texture - so you can get pretty good results with a fairly simple algorithm.

hth
Jack

Thanks, that's pretty helpful. Is perlin noise difficult to do real time?

Quote:Original post by Battagline
Thanks, that's pretty helpful. Is perlin noise difficult to do real time?

No, not really. As I said - you can use the hardware texture filtering to "fake" more information than you have (provided blurry is ok!).

I'd hazard a guess that you could get good results from a 32x32 source texture - calculating 1024 values every couple of frames (or morphing accordingly) probably won't hurt performance too much.

This webpage has a very good explanation of the algorithm. It goes into a lot of detail on interpolation - but that is the part I'm talking about using the GPU to do for you [smile] so you can probably ignore it.

Also, the following from the Graphics Programming and Theory forum FAQ:
Perlin noise tutorial: Fast rendering of (faked) multi-octave Perlin noise textures (by Hugo Elias).
Improving Noise: Paper presenting the new Perlin noise reference implementation (by Ken Perlin).
Perlin noise reference implementation (source code).
Making noise: The history of Perlin noise.

Should be some good reading for you [grin]

Jack

That last post was me. I wasn't logged in I guess. Anyway thanks again.

:)