Is it better to separate the head for the facial animations and lip sync? Also is it better for the performance reasons?
In what I’ve seen, you can have a single character mesh for the whole rig, but if you want to change the head, you have to change the whole body too. Commonly, you’ll see a single base body used for lots of characters but the head mesh (snipped around the shoulders) can be swapped out to create a large range of characters. So, separating the head from the body is more for production efficiency rather than performance efficiency reasons. You’ll still share the same underlying skeleton / rig on body head and body meshes. If you don’t need to swap out heads, it’s perfectly fine to use a single mesh.
One other thing to note: You are limited to 8k texture sizes, so if you have a full body mesh, you need a texture with enough resolution to cover the whole mesh. I haven’t had any issues with 8k textures not being high enough resolution, but… if you really need fine detail in the skin textures, you could break the body up into separate skelmeshes so that you can use multiple 8k textures. Keep in mind, that’s probably going to eat up a lot of VRAM though, so it’s usually not ideal for real time games. If you do want to do this for a game, you’d have to be careful about your LOD’s (though, who knows if that’s going to be relevant for UE5).
What about other things like blending?
This:"
Why would i have a problem with the texture size? My npc mesh has 8 materials to which i can add high res texture if needed.
And 8k texture - i think that’s unnecessary.
Btw, this is for a game.
Are you using clusters or morphs?
Morph wise only the vertex containing displacement is recorded so data footprint wise the performance envelope is no less or greater between using full body or just the head. Performance wise vertex animation was updated from being CPU bound to GPU based in, I believe, 4.15 so even though there is always a performance hit when resources are used the result is well with in the expected performance curve as to usability.
I have a dialogue system to which i can add voices and animations. I want to make dialogue similar to Skyrim.
So what is better, clusters or morphs?
Well the stock answer is “It depends”
Morphs
If the character is required to “act” as a hero then morphs would be the better option but there are rules that are required to be followed to get the correct result but the fidelity is hard to beat.
The number of vertices needs to match the target else the blend will not work.
The blend targets needs to be included with the mesh export.
Characters require their own targets.
Animation can not be shared unless they contain the same target names.
Clusters
If the character is an NPC then clusters would be a good option.
Clusters can give you real good results if you have enough joints and can be animated with in the same channel as the rest of the animation migration with in the animBP. Clusters can also share animations between different characters and the joints can be added to the rigs hierarchy as children.
You can also combine both if you wish but as important as the two options are how the asset is set up determines which of the two choices are best. What applications or APIs used to author the animations, at the moment, makes a difference as to choice based on tool availability