Page 1 of 1

Better emissive blending

Posted: 19.11.2020, 05:19
by cubicApoc
Anyone who's tried using NightTextures to make planets glow knows that past a certain point, you just can't make it look good. The results often look something like this.

The transition from day to night is too sharp, and diffuse shading darkens the texture well before the terminator. While this is fine for cities or subtle volcanism, it doesn't work as well for hot Jupiters or lava worlds. Here, I've replaced the planet's surface with a high-poly CMOD. The day texture is in the diffuse slot, and the night texture is in the emissive slot.

Now, the night texture is drawn over the entire surface and doesn't fall off toward the day side. This is what a red-hot planet should look like. These textures are just placeholders, but you get the idea.
It does have some issues, though. If you define an atmosphere, there will be tiny gaps between it and the surface. These aren't really an issue from space, but you'll see them if you land. It can't do anything other meshes can't, so no eclipse shadows. I haven't done much in the way of testing this, but if you want to try it yourself here's a template you can use. Open it in a text editor, switch out the textures, and generate tangents in cmodview if you added a normal map.
(2.13 MiB) Downloaded 398 times

Posted: 07.09.2023, 19:05
by Eric Nelson
Not to be critical or anything, but the poly resolution looks anything but excellent.
The cmod models are 2 subdivision levels shy of being excellent in for planetary modeling.
Though they're better than cmod models that are even just 1 subdivision level lower.
Though this subdivision level looks good for 3DS meshes, from my understanding, cmod meshes can have very high resolutions by comparison.
Though I appreciate your effort on the model.

Posted: 07.09.2023, 20:31
by MrSpace43
Bruh stop necroposting. You've literally already done it twice today.

Posted: 08.09.2023, 02:07
by Eric Nelson
Well SevenSpheres' showed me a link to this to reply after my honest review on the Celestia GitHub site about the emissive models on GurrenLagann's addons.

Posted: 08.09.2023, 03:14
by MrSpace43
That doesn't mean that you need to post a reply on this topic, especially because that there has not been a new reply on this topic that was made nearly 3 years ago.

Posted: 08.09.2023, 03:17
by SevenSpheres
I didn't really "direct you here", I just said this is the source of the models, but this is definitely a more appropriate place for the discussion. You'll have to wait for a response from cubic, but I don't think there's much need to modify the mesh to remove small imperfections only noticed by one person.

Close up vs afar

Posted: 08.09.2023, 17:54
by Eric Nelson
While as you said there are no perfections when it comes to making a "sphere" on stuff like Celestia, or common computer software (due to the poly nature and generally computer images being made up of pixels), there is the matter of what looks excellent even from a close up vantage point (not to mention PCs are getting better each time).
It's just that Cubic's resolution is 2 levels below what many consider the magic level for approximating spheres of celestial scales at practically any distance and performing smoothly on many PCs at the same time, though Cubic's level isn't really bad (and I admire the effort he put), but I described the quality as average, since it looks good from orbit, but not enough of uniform quality from surface view or even the cruising altitude of aircraft, unlike adding 2 more subdivisions, which would look excellent even from the surface and still perform smoothly, and also create a value that's a power of 2 divisible by 2 (adding to the benefits).
It's actually a huge difference (4 fold in outline and 8 fold in number of triangular faces) to go 2 levels higher, to be honest, not really small like you (SevenSpheres) mentioned.
And such "imperfections" are noticed by multiple people, pretty much anyone who uses it often and especially pans the view while on the surface.
Thanks to you Seven Spheres for letting me understand what was going on (and I updated my earlier comment stating that you showed me the link to here).

Posted: 09.09.2023, 00:52
by SevenSpheres
Eric Nelson wrote:And such "imperfections" are noticed by multiple people, pretty much anyone who uses it often and especially pans the view while on the surface.

There's no evidence that this is the case. You were the first person to mention it, and before you did I hadn't noticed and assumed the model was (near-)perfectly spherical. I had to land near the poles and move around to see it.

Why did you report my post? It wasn't meant as a personal attack, just an observation that as far as I'm aware, you're the only person who's noticed this.

Posted: 09.09.2023, 01:52
by Eric Nelson
I didn't say the model was perfectly spherical (which's impossible for a poly model anyway), I just said it can use as much as 2 more subdivision levels to make a great appearance for simulating planets and moons.
And I wasn't trying to promote a personal attack either, but I just wanted to point out the reality of what is an ideal resolution.
The report was due to the fact that when you said I was the only one who noticed the "imperfections", I was implying that you didn't take into account that anyone can notice them if they check clearly, and that goes for anything just to clarify, though I accept stuff like Celestia for what it is.
Now the subdivision level the emissive mesh is at is good from a view where you can see it as a whole, but Celestia lets you land on life-sized celestial bodies, making a high resolution necessary.
Not to mention that a power of 2 divisible by 2 means numbers such as 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072, 262144, 524288, 1048576, 2097152, 4194304, 8388608, etc, and that 8 is the recommended level for such a mesh.
It seems like I keep causing problems when I'm not trying to and I always check what I'm typing before posting.

Added after 5 hours 13 minutes:
Here's an example to explain it all.
Cubic's emissive planet models should serve as a reminder of how we view things depending on distances and judgement.
What looks like such from afar or at quick glance can be a different matter up close.
Not to mention how profound of a difference of 1 or 2 levels is, not to mention that using a power of 2 is the most ideal useful way to render a mesh in good quality, meaning a subdivision level of 8 follows all that.
The last 6 screenshots are from Blender, using normal mode for those who can detect the patterns well and edit mode for those might not see the patterns at first.