Nanite doesn’t support translucency or masking so no foliage that uses those (pretty much all of them) won’t work. This may change in the future but for now it’s still good practice to create billboards for trees.
Megascans trees use Octahedral Impostors - they have masked leaves - I am wondering if they will release some ultra-high poly trees with the leaves modeled too but I have a feeling that even with nanite the polycount may be too high - will be interesting to see!
I’ve experimented with creating nanite opaque trees and I was able to get around 60,000 at 20fps (nothing else in the scene), but I wasn’t happy with the shape of the trees in the distance.
Yes, old school cards - Octahedral Impostors (dual axis cards with inter-frame smoothing and custom lighting) or simple billboards. At a distance they look better than nanite meshes in my opinion - and of course they are much faster. I’ve create a scene with 6 million billboards at 20fps running on the same hardware as above.
hey thank you. This makes sense. Hmmm… do you have a link to your test?
Would be cool to see more from it.
But heck, that basically means… we could use high poly evermotion trees… just for example.
The downside, non of the trees will work with wind modifiers.
Another issue… Megascan trees would allow you to change the attributes of the foliage… like fall, etc.
However… it looks like the solution is to use actualy 3d geometry and convert it all into Nanite. In the tech demo they also spoke about, that tesellation is not supported anymore. Meaning, forget about displacement maps and rather load the high-resolution mesh, convert it to Nanite and that’s faster and cheaper.
Soooo, I am assuming that’s the way to go. Just not sure what we could do with dynamics, like wind.
Yeah the potential is great, but yes things like the wind, season and tree health (with splat-maps) systems that are coming soon are more aimed at LOD based trees.
There is also the issue of foliage leaves and their lighting - the way it’s done now is with two-sided foliage shaders that calculate the amount of light that passes through the leaves - I doubt nanite will be able to support that at the moment.
I don’t think that nanite trees will be a reality for a while yet.
The way I’m doing it in my game is to use LODs for all trees and focus nanite on buildings and props. I mix the Screen Sizes of the tree LODs and I also include some trees without billboard LODs to remove any seams and help with a better wind effect.
that makes sense. You know, I was thinking… when looking at the City Sample.
If you have trees in a close-up shot and you want to have high fidelity… it might be best to use a bunch of Nanite Trees… but maybe mix in a few non-nanite with the wind.
But… it might just look very odd, if some trees are stiff as hell, while others are showing the wind effect. Plus, the fidelity difference.
To make sure, that all the trees kinda have the same visual fidelity, and still are cheap for runtime. Heck… this is not an easy task.
I haven’t heard any new news about their trees - The last thing I read was that they were going to go around the world capturing trees - but it’s possible that had to be postponed due to current world issues? Plus each tree would need to be captured in each season. Their trees will be built to support all platforms and engines so I don’t know what their focus will be on nanite but I can imagine they are very interested in it…
They almost look like they are still using LODs - did you save it with “Use Nanite” - hovering over the mesh should give something like: (although this is fairly low poly)
Nanite uses edge collapse decimation to simplify meshes. It will not work on trees, because most of the geometry is defining the silhouette and can’t be collapsed without destroying the shape.
I don’t understand what you’re asking here? A tree converted to Nanite will behave the same as other Nanite meshes. The issue is that the strategy nanite uses to reduce geometry does not work on trees because the geometry in it is defining the silhouette