Curvature artefacts

copenhagenjazz

Hi

I am trying my hands out on Substance Designer trying to bake out all maps in Substance Designer. However I am getting some artefacts with the curvature map I try to bake out.



Now these artefacts in my curvature is certainly caused by the gradients in my normalmap. Is there any way to get the curvature generator read my normalmap in the same correct manner as the renderer or will I have to work around this issue by splitting the UV's and the smoothing groups?

Any suggestions or should I just accept these limitations?

Thanks

Jerc

could you try lowering the curvature intensity (details)? It looks very saturated at the moment and it may cause unwanted detail to come out. You may be able to get a better overall look with smoother variations, with a lower value.

copenhagenjazz

Hi Jerc

Thanks for your suggestion I have lowered it and indeed it looks a lot better. I intentionally exaggerated the settings here to illustrate the issue.

Lowering the intensity however is not really my preferred solution. Although I love baking everything in Substance I cannot unsee an artefact if it has presented itself :P So I would rather solve it somehow rather than just covering it up

Jerc

Would you be able to share your low poly + normal map for us to take a look at?

oblomov

I don't think the maths we use for combining the curvature coming from the discrete geometry and the curvature coming from the normal map are precise enough for them to perfectly compensate each other in such a case. I suppose the only way to get a perfectly flat curvature for hard surfaces with such normals would be to directly compute the curvature on the high poly mesh instead of the low poly + normals, which would require a new baker.

CheeseOnToast

Please do this, if it's at all feasible. I assumed that's what you were doing in the first place and couldn't account for issues like the above, and the thin black seams you get at some UV borders (even with padding).

Would generating from a world space/object space map give better results? Essentially taking the low poly normals out of the equation?

copenhagenjazz

Jerc I just sent you a PM with the example files.

Thanks for looking into this

copenhagenjazz

Hi Jerc, did you get the files?