I wonder, how do they look from the ground? Is it a lot better than regular “fake” clouds or it’s not worth it? Also have you thought about using regular spheres instead of concentric spheres and then “bake” the detail into the outer sphere? not sure if it would affect the quality much but would improve performance.
They look good, but I fade them out real quick in favour of a cloud plane which is cheaper and almost as good.
The clouds look spectacular when mountains are involved:
One of the bonuses of the post-processing is all of that banding goes away.
You can try it out on windows if you like: MEGA
wow thanks for the vid and demo. It ran fairly well on my gtx 960 with a few hitches when flying straight down or up, I believe it has to do with the loading of higher landscape lods not very concerning. (can’t say fps for sure, the default console key never works with me, I usually remap it in my project)
the clouds look amazing, they look even better when moving, how did you achieve that movement/scattering effect? Speaking of effects I did notice some weird lines while moving through them, sort of noiseish…is this the banding effect you talked about?
The movement/scattering is unintentional and a side-effect of using a low repetition projection mesh for the clouds to display on, so it kind of “boils” as you move towards or away from it. Sometimes it works, but it’s kind of annoying. It’s better to have the clouds mutate over time rather than with movement.
The banding is soft, but there’s some other artefacts as well to do with the mesh being weird and the noise occasionally doing strange stuff around zero locations.
Maybe a hybrid approach would be the best bet…
Thanks for the positive feedback anyway 
Oh as for the water changing color, that’s what I call a “unity bug”, I call that to all weird rendering issues. I have that problem with the ocean project as well, I think it might have to do with screen space reflections (if you move the camera so the sky is not visible water has nothing to reflect so it gets dark) I haven’t confirmed it yet…now with planar reflection I could test to see if it solves the problem.
Did the water in the demo change colour for you? A while back I had some problems with reflections when the water was a long way from the world origin.
Oh right, yeah typical SSR problems. That’s only to be expected, I think most people are used to it by now.
Yes it did a bit, also the world seems to get brighter and darker sometimes in a unnatural way when very close to ground not sure why it happens. but above the clouds everything looks so beautiful.
It’s not so much of a color change, more like a shadow being cast on it, the reflections get darker or pitch black…it’s defenitly the same issue as the ocean project, as the camera rotates in specific angles the dark shades move accordingly, I am pretty sure it’s related to SSR, hopefuly planar reflections will solve those issues I have been tackling with. I can make it go away with a post-process effect that turns pitch black into water color…but that is only an half solution.
Right, midway through the rewrite. It is now a ray-marched 3D procedural cloud renderer.
ShareX capturing is causing the jerkiness - I’m getting around 30fps on average at about 720p.
I haven’t done anything about depth sorting, the most opaque sample owns that pixel and is responsible for lighting it. I’m going to have to add an input for the sun direction, since shaders have no access to that by default, which is quite annoying Epic. 
After that I can light the clouds when the sun is behind them.
The controls are things like storminess, fluffiness, reflectiveness, etc.
Performance is about the same but the quality is through the roof. This might become a next-gen effect if I can’t fix that up.
Glad to see that you kept working on this project!
Nice fluffy clouds 
Looking forward to the next updates.
Haven’t touched it in a while! I was running into too many difficulties with view space transforms on the post-process version being weird. I’ll revert to the hull method probably as it solves a lot of problems and in the end there was no real speed gain from the post-process method. Or maybe a hybrid method.
Roger!
When are you starting the gameplay side?
Already have a precise idea on how you want it to be played?