Reply To: Polybook – Jake
@louhikarme Composing the nebulas in Eevee is going to be fun. Unfortunately the final render itself is going to have to be Cycles; Eevee is a no-go for this one it turns out. On the other hand, much is so similar between Eevee and Cycles now that whatever I compose in the former should work out well in the latter. It’s not perfectly one-for-one; when making those spotlights for instance, there was a very marked difference between how they looked between the two; in Cycles the lights had to be much smaller and moved much closer to the hull to look the same as they did in LookDev mode (which is Eevee), and as you can imagine this took a whole lot of render-regioning and adjusting over and over again – it was time-consuming, and although I got the lights exactly how I want them in Cycles, unfortunately in LookDev they don’t look nearly as great anymore.
Why these differences? And why is Eevee a no-go for this project? Well it’s both for the same reason, and it’s really my fault unfortunately. The problem is the scale of the ship. Our CGC instructors always tell us how important it is to build to the correct scale. It’s not something I neglected per-se; the ship is built exactly to the measurements defined by the plans I used, including the correct scale. The problem – lol – is that those plans are for a 100-inch-long 1:120 physical studio filming model. So my great big starship is actually only just over 8 feet long. Some of those details I modeled are less than a millimeter long or thick.
Normally this wouldn’t be a serious problem; if they could take an 8-foot practical model and make it look the right size on a movie screen I shouldn’t have any trouble doing the same thing. That was my thinking when I started, and it’s still valid to a point. What I didn’t realize then, but I have learned since, is that Eevee’s lighting system has serious trouble working as well at micro scales like this. Take these early lighting tests I did in both Eevee and Cycles:
Look at how much light just seems to be spilling right through what should be solid objects and illuminating surfaces it shouldn’t in the Eevee renders. I’ve tried everything to remedy this in-place, including adding thickness to some surfaces which didn’t have it, and adding more mesh objects around the light with the hope of occluding it, sort of like making light-boxes to prevent this kind of spillage on physical models. Nothing worked. Researching this problem, I learned that apparently if a surface is thin beyond a certain threshold, Eevee just can’t “see it” for the purpose of deciding whether it should be casting a shadow. The only fix is to add more thickness to the object or scale your whole scene up until the mesh is “thick” enough to start occluding light. Again, this is only a problem you’d ever encounter when working at very very small scales, like I unfortunately am with the details of this model.
Obviously adding thickness won’t work for me; there’s only so much I can add before things start visibly clipping through other things, and it’s not enough to actually fix the lighting problem, so it’s just needlessly adding to the poly count. Scaling the whole thing up would probably solve my problem, but I’m actually afraid to do that; scaling things up will almost certainly play havok with some of my modifiers, especially Booleans, of which there are multitudes. Plus the project is so big at the moment, in terms of the memory size, number of objects, the poly count, and other factors. Making all the invisible stuff visible so that I can scale it will add enormously to the already huge poly count, and then trying to scale it all up at once….well, my poor old computer might get right up off the stand and walk out on me. And that’s before I get into objects that don’t scale the way mesh does, like light objects, which I’d inevitably have to fix by hand. Not only repositioning but redoing all the intensity, size, radius settings, etc. I don’t want to take the risk or invest the time and effort this close to the challenge deadline. So Cycles it shall be!
That said, I do have some post-challenge plans for this project. But I’ll get into those later.