Hi, I’ve been self teaching myself a ton of stuff in UE4/Blender/Other and pouring through documentation galore and still have a lot of questions on some things and thought I would start with this.
In dealing with geometry on a massive scale such as a tree that is many kilometers wide and reaches past clouds/into the skybox, are there any best practices or requirements for modeling and texturing this as a primarily non-interactive background set-piece/cinematic visual?
I am currently considering modeling the roots/stump area and then making a series of meshes that are easily modified to stack on top of each other to comprise the trunk. I would use tiled materials for most of it sprinkling in some custom ones where applicable.
The size of the polygons are also a talking point for me as I know that the distance which this will be seen from is relatively far and I should be able to drastically reduce the mesh detail and make up for it with some material tom-foolery but I am unsure as to how far I can push this.
The closest the player would be to this mesh would be near the roots looking up out of a vertical cave/canyon.
Basically I’d just like to get some general guidance dealing with large scale meshes before I try to fumble my way around it.
Interesting question. I’m also still learning a lot of stuff so I might not have the best answer, but I think what is of concern is mainly the number of vertices that are on screen.
So If it takes up half the view (even if its really far away), that should use more or less the number of vertices you use for a character, or any other mesh that’s near by, that takes up the same amount of screen space. Same goes for texture resolution I suppose.
Then as you get closer, you can use higher res LoD meshes, and unless you plan to go flying around the whole thing I think you could maybe get away with just adding more detail near the bottom of the nearest LoD mesh. Otherwise cutting it up into lots of pieces, each with their own LoDs is probably a good idea.
What I’m wondering about is how much of the mesh is still processed by draw calls if you’re only looking at a little piece of it…
Thanks for the reply. I expect the best way for me to figure it out is to experiment eventually. I haven’t gotten around to learning much about LODs yet barring particle effects and I figured that would be pretty relevant to my needs here. In-game the player would only see a portion of it so I would most likely remove most of the unseen geometry during game-play. I think I would be better off pre-rendering a fly-by for cinematic type purposes.
Another point to look into is visual trickery. If you pop open some of the epic demo projects, racing for example, you will note how they use forced perspective in order to trick the player into thinking that something is large and far away when really its a lot closer and smaller than it appears; the pyramid on the track is an example of this.
I’ll definitely take a look at these examples soon because this sounds like it would really help me out. The view of this asset would be from a generally specific angle during gameplay and only near the end of the main story experience barring a fly-by in an opening sequence that will most likely be pre-rendered so if I can figure out a way to use perspective for in-game rendering it would save a lot of performance. The majority of the game takes place in an underground “vertical” style cave/ravine so the only time it could be seen is near the entrance at the final sequence I have planned.