Procedural Coral
Here you can see a mild amount of UV distortion, one of the risks of running a procedural texture. It likely won't carry over, and is fixable if it does.
In fact, I might just switch to 3D noise and object coordinates, and bake that...
I had to *fight* Sketchfab to get this thing to upload. Throwing out UVs, flat out ignoring them, breaking textures... their new upload system is garbage. So I guarantee nothing.
Basic turntable rotation through cycled parameters on procedural coral material
Demonstration of a cookie-cutter boolean—in this case, I mirrored cube—turning a single piece of coral into uncountable possible smaller pieces.
An anaglyphic turntable tour of my model, which I'm very satisfied with... seeing your procedural material pop out at you is a unique kind of fun.
Let me start by saying that producing a procedural material does not mean that you don't need to know how to do it yourself. In fact, you need to know how to do it so well, you can explain it, piece-wise, to your machine. That's what this came down to.
The coral began as a sequence of metaballs, or mathematical functions describing a 3D surface. If I was leaving this in Blender, I might've even hollowed it out and left it that way for the possible dynamics. However, since I'm exporting to a few other pieces of software, I like to reduce it to meshes and, if necessary, armatures and curves. There actually aren't any shape keys involved here, but the material itself is clearly animated.
The four parameters I decided to expose, and make flexible, are the two colors of the coral, the lacunarity of the gradient noise that makes up the surface, and the scale of that noise. The other factors involved were more likely to break the model than beautify it, so I stopped there. It should be relatively easy to replicate in any other visual shader editor.
The breaks in the second video came from a "cookie cutter" mesh—or at least that's what I call it—attached to a boolean difference modifier. The great thing about procedural textures like the noise in this material is that they tend to be self-healing if you work them right—there are no pre-defined UVs for the cut, but figures them out on the basis of my node graph. You can actually see them animate a little, but since that's not a feature I intend to export to the final rendering engine I'm not concerned about it.
By simply shifting the cookie cutter, and applying modifiers, I can readily produce a whole array of different pieces of coral from the same base mesh. (Calculating, we have 2⁴⁸ different possible color combinations plus full flexibility on scale and lacunarity; but if this ends up also being procedural in the game—which is a little up-in-the-air, as they say, for me right now—then I'll likely only have it pick from a selection of that, ensuring that it neither breaks the material or the motif of the game.) While there are definitely uses for everything cycled through here, my goal for the game is beautiful-but-dangerous, and I'll likely have a stricter rule set.
The final model was only about 200 faces, if you can believe that. Proper texture baking works miracles. I may end up planar-decimating the cut pieces, due to the possibility that that flat surface with a single hole in it may go totally-over-the-top on its face count, but then, I doubt it will be significant.
Development time is basically since sometime yesterday evening to around 8:00AM this morning, I'm going to roughly estimate that, not counting rendering time, it took me two and a half hours.