I agree. A lot of people don’t want to get squeezed out of a good product, just because it’s ridiculously high poly…
all the time i am grabbing scans from places like sketchfab with intent to retopo them, or just use the thing as a general base to work a new model from.
i think the key is just that, i know I’ll need to do that.
if something is billed as being “game-ready” when it really is not, of course that is a problem. Right now, I suppose the general public doens’t really know exactly what you can get away with because of nanite.
Perhaps in the future, more general guidelines based on targeted platforms might exist, and then somebody could make more informed decisions about “will this scan be viable in my game, or would i need to do some manual processing on it?”
For me, based on my limited knowledge, I think the answer to “how much triangles does a nanite model need?” is simple:
only as many as needed to get the desired silhouette and surface details, and not more.
its not really different answer from the question, “what resolution does my normal map need to be?” or “how many tris for a traditional game character?”
the answer is always “minimum needed to get desired ________ effect.”
For a traditional character with normal map, the geometry defines silhouette and normal map defines surface details, so nanite just combines those two needs into one.
Any model could pretty quickly just be run through a decimater to find the min resolution needed to hold onto the desired surface details. And then that can be measured against traiditional low-poly + normal map package.
It sounds like epic has of course already confirmed that nanite models are going to be more efficient for many reasons, so personally I’d just make my scanned models as efficient as possible and trust the experts from there.
2 Likes
I’m working on a new package for Marketplace. I’m still stuck in the 80’s and poly counts over 100K make me nervous. Just curious on what the poly count threshold is recommended for a mesh to switch between nanite and standard mesh? In a similar vein, at what point does one need to consider adding LOD meshes for standard meshes given the improvements in rendering speeds?
Was wondering why it is so important for people to have a good topo (before Nanite) if they just want to use the models beside the compressed size.
If you have a dense area at let say at an broken edge it is used to have less noticable texture streching (and of course details).
For myself it would be
- compressed size
- vertex paint (you need a clean topo for good results)
- redo the uv´s (that is why i have and use a lower res mesh)
- no normal map (or if possible just vertex paint without textures)
Still testing poly paint this would be great to use without textures in terms of gpu load. But i still just cant really figure out how Nanite display the verts 
and so painting often get just destroyed.
Would be nice to know because i still test how to use it a better way beside just build a mesh even if this is possible.
1 Like
I think ‘whine’ was the correct term. 
Goot topo is most noticable while editing a model (the ease of adding edge loops and detail where it counts with a few clicks) but also animation. You can’t properly animate shoulder rotation or leg bending if the topology is bad, you’d get stretched and pointy shapes all over the place.
2 Likes
Just to be clear. I was asking only about SM and not SK meshes. They are very different. As far as i know Nanite does not support SK at the moment.
I dont think anybody would use a million poly SK mesh, and of course you are right, you would need a very good and clear topo for a rigged model.
In your case seems like you want to expand and modify models so then it is also of course much easier to work with good topo.
That said, it would not help just do a simple decimation with ZBrush to bring the compressed size down, the results are good for this but the topo is almost destroyed after doing this.
And i watched a few older videos on youtube about Nanite. They clearly say that they want to push the limits with all this very high poly meshes but it will also be sometimes a better idea to use lower polys because of the size.
1 Like
preach! glad im not the only one - so far my experience with Nanite has just been a lot of work to try and optimize other peoples’ non-optimized content. certainly feels a like ‘too much, too fast’
1 Like