This is so perfect for my Planet Update!! But I need money to even buy it. I hope I will get this asset in the future. How does it even work? Good work btw!
And also one more question, are you planning to make O`Nielās atmosphere or Eric Brunetonās Precomputed atmospheric scattering? These atmospheres are more realistic than your custom atmosphere.
@foxyspace
Great!
I will make a tutorial video for the new planet system so itās clearer how to configure it.
If you want super realistic atmosphere you should use HDRP, which comes with its own planetary atmosphere system with a trillion settings. The current atmosphere is designed to be really easy to configure for any colors you want, and this is achieved using one sample, which is pretty impressive. That said, I do plan to improve the color calculation code and settings to make it a bit better.
Itās the same thing with the atmosphere, Iām looking for a technique that is easy to configure and has good performance, not necessarily what is the most realistic.
1 Like
I donāt know why HDRP mode switching scene or running mode exit often black screen, have to restart to restore normal.
@YangYuanJie
Does this issue occur in a new project with just SGT? If so, let me know the exact version, and what steps I must take to replicate this.
Hereās some more progress on the volumetric clouds:
Itās not super realistic yet, but I think itās already interesting enough to include in the next version. Itās part of the sky feature, so you pretty much just enable it in the sky material if you want.
I also made cloudsphere dual sided, and Iāll include that as well.
4 Likes
I love this ā it would work great for my use. Are you planning on including both cloud approaches youāve been developing?
That looks great! Have you tested it on an Earth sized / really large planet yet?
Looking nice, Would it be possible to set the cloud layer height and maybe have multiple cloud layers at different heights? Or is that outside the scope of development at the moment?
@rezuma
I donāt want to maintain lots of different but similar systems so Iāll probably only include the best. I donāt think the particle based approach is feasible as the overdraw is just too high and basically not possible to overcome without drastically decreasing the sense of cloud density. It works similarly to the infinite starfield cloud demo scenes in SGT, which also slow down if you add too many particles.
2 Likes
@ZhavShaw
Not yet, but itās based on SDF volumes so it should work with any size planet.
1 Like
@RetnuhStudio
You can change the height and thickness and coverage etc, but only one layer.
However, I just thought of a new approach that may be able to support multiple layers and run faster than this, Iāll have to experiment with it next week. Stay tuned 
5 Likes
You are faster then Space Engine!!
@Darkcoder1
Do you plan to implement multiple cloud solutions?
A high performant particle one for mobiles and a volumetric for high perf. PC?
Or do you think the volumetric approach could be performant enough for mobile as well?
Second, is it feasible to use the cloudsphere alpha values to determine the spawn areas of volumetric / particle clouds?
So we have a good looking macro cloud structure, that blends over to Detail clouds?
Do you happen to have any solutions for working with large scenes in the Editor? In-game Iāve got my own origin shifting solution, but Iām afraid to do origin shifting in the Editor.
And what about fake erosion? You might find this video interesting
@JamesTCookie
The existing (but improved) cloudsphere will be the low end fallback, because itās so simple.
I donāt see any scenario where the particle system is the low end fallback, because the fill rate requirements are too high, and mobiles especially seem to be bad at this.
The particle approach used the cloudsphere cubemap as an alpha source, because itās mostly one sample per vertex thus fast. However, ray marching requires a lot of samples per pixel and sampling a cube map is too slow. To make it work, a slice of the cube map would need to be rendered into a 2d or 3d texture that is then marched through.
This new idea I have would be a replacement for cloudsphere and would work with the cloud cubemap and possibly manually placed cloud stamps. I still havenāt written much of it so itās mostly theoretical though, we will see.
@ZhavShaw
Nope, and since Unity keeps changing everything every year I have little desire to make any scene tools.
For massive scenes I prefer the idea of procedurally spawning everything and only using the editor to configure each thing that will then be procedurally spawned.
2 Likes
@vistaero
Super cool video!
However, I much prefer the Planet Forge approach of using pre-baked heightmaps and combining them together. This way you can use super complex erosion techniques and even real life DEM data with no difference in performance. The main downside is it requires a lot more memory, but these days thatās not really an issue.
Also, most importantly for me it means I donāt have to spend months coming up with various terrain algorithms. SGT has the TerrainSimplex component which uses Unityās own snoise function, but I found there is a bug somewhere in that that causes value spikes for one in a million float values which is so annoying.
Also, you can just use Gaea to create believable macro erosion heightmaps, and also smaller high frequency detail maps. Itās a great tool!