the work im coming from is as an environment artist dealing with blended tiling sets that used to have the benefit of tessellation. now nanite says, treat it all as a zbrush prop without the baking. i trade time for quality because i now have to sculpt everything as apposed to having a non destructive work flow of vertex blending my details
Nanite static meshes will support everything which is possible with UE4 static meshes, including Tessellation, vertex painting (they were looking for a different name for that ) & even non-rigid meshes.
They are experimenting(or will) with Nanite for foliage
Landscapes will get Lumen support
Landscapes might not use Nanite & suggested to use Virtual HeightField Mesh.
There’s another LiveStream in two weeks on Nanite, so lets ask all these questions there.
How can Nanite replace tessellation if it they are almost completely orthogonal?
Nanite doesn’t support translucency, displacement and a lot of other useful things (yet). Tessellation was a simple abstract operation, it’s not just for one use case that Nanite is.
Now I can’t draw some things with maximum efficiency. When I move to UE5, I think, I’ll dig source changes and will try to bring it back. Maybe someone wants to collaborate on this?
How do we apply any of this to SkeletalMesh? My characters now look flat and smooth. I think nanite is not a good substitue for displacement, please put displacement back in UE5,
Please do not depricate tessellation! I cannot afford for my project to be dozens or even a hundred of gigabytes! I’m relying hardly on texture tiling for that matter. Also, I can’t make working Nanite on my machine, don’t know if it requires dx12, or is there something missing in a clear UE5 project that I have to enable/download or something.
Any update on this? I haven’t seen any alternative to tesselation so far or any talk about it.
Currently Virtual heightfield is a complete mess and isn’t even close to an alternative
NONE Of the proposed alternatives are even viable for Procedural Meshes. Our game is using them and now is going to be stuck in 4.26 forever just because tessellation has been deprecated.
For our game this completely breaks visuals that we were going after. Sad that Unreal deprecates something before it has a suitable replacement.
I wonder what will they deprecate next… LODs? “with Nanite you don’t need LODs!”, “right, you can’t use Nanite with skeletal meshes, but with Nanite you don’t need LODs!”
What you’re thinking of is deprecating manual LODs. Nanite still uses LODs, they’re just generated automatically in a way that (1) makes the transitions (virtually) imperceptible, and (2) allows them to be streamed in & out the same way you would a virtual texture (hence the term “Virtualized Geometry”).
I wonder how it’s possible to do good enough automatic LODs so that they’re not cutting out important elements, turning square elements into triangles and so on.
It certainly is possible, with static meshes at least. Nanite shows that. I would not be surprised if something similar was possible with skeletal meshes in future too.
I understand Epic’s choice to deprecate tessellation in favor of Nanite. It makes sense for high detail objects.
The problem is that it omitted some other use cases for tessellation, such as landscapes. It’s not feasible to import large terrain mesh with similar up close detail to landscape material with tessellation. It would have multiple terabytes and would eat up all the memory.
That being said, my hunch is that since the motto of UE5 seems to be “virtualize ALL the things!!!” (Virtual textures, virtual geometry, virtual shadowmaps, virtual lightmaps), I would not be surprised if virtual landscapes were next on the list. We already got virtual heightfields, so it’s just one step from there to make landscapes use them natively. They’d provide both better quality as well as better performance than tessellated landscapes.