4
$\begingroup$

When I render my model in Cycles (I applied a dummy material just to check), there are some bad/glitched shadows in some spots.

Like this:

enter image description here

But, in shaded view (smooth shading applied to the whole mesh), it seems fine:

enter image description here

I checked the face normals, and they're fine, vertex normals look fine too:

enter image description here

This is a retopo of a sculpt I've made, it might not be perfect but there are no triangles or n-gons. It's also strange because those darker spots are triangular. I also checked if there were two faces one over the other, but there are none. Everything looks fine.

I don't know what may be causing this, any help?

I think I should point out that this happens on more than one spot and only when the spot is in some degree of shadow. If it's lit it looks fine.

$\endgroup$
  • $\begingroup$ Coupld you upload the .blend file to a site like pasteall.org? Looking at the picture I would agree with stacker's answer. It looks like you've got some really wacky normals: normals coming out of edges, multiple normals on a face, normals pointing peculiar directions, normals that are not centered in a face, etc. The last of these is usually caused by an n-gon that looks like a quad (a “quad” with a subdivided edge. $\endgroup$ – PGmath Nov 21 '14 at 16:11
4
$\begingroup$

Those artifacts are showing up for a number of reasons. The first is that you have polygons that are very non-planar and this will usually cause artifacts for lighting.

Here is an example that shows a quad but it's folded sharply and two distinct triangles can be seen when viewed along the plane of the quad. Try and make sure to keep quads as flat as possible. Non-planer quads will also give you problems with texture mapping so it's something to always look for and repair.

enter image description here

Another problem you are having is that you are rendering with a very low polygon model. Cycles has difficulty dealing with models which are not subdivided at least a couple times. It's best to use a subdivision surface (subsurf) modifier to handle this since it can be removed without affecting the base model or it can be applied so the model will be permanently sub-divided. Here is a sequence of images that show the effect that various levels of subdivision can have on rendering artifacts.

enter image description here

Another way to deal with these rendering issues is to use more lights. Global Illumination will also help since these areas will receive indirect light that will also help smooth things out. The following image shows the effect of changing the angle of a single light source in order to show how the area is affected by direct illumination. enter image description here

For the final image here, I did not alter the mesh at all but instead only relied on the Sub-Division Modifier at 1 iteration and also there are 3 lights as well as an environment map. The ground also adds a bit of indirect light that helps to fill in the lighting for areas that the other lighting is not reaching. This all adds up to a mesh that looks pretty decent even though the bash mesh is low poly and it has some fairly substantial non-planer errors. If the Sub-D modifier is increased to use 2 iterations; the few problem areas that are still showing should no longer be perceptible.

enter image description here

$\endgroup$
  • $\begingroup$ I was keeping it low poly on purpose. I will try to fix the non-planar faces, I really didn't notice that. I didn't even know it was possible to have something like that, really. Is there a fast way to find such faces? $\endgroup$ – Paul Nov 21 '14 at 22:37
  • $\begingroup$ I'm not aware of any tool that can find non-planer poly's directly. However, if you Triangulate the model then use the Tris To Quads tool, you will see that very non-planer polys will tend to remain as triangles instead of being converted back to quads. Unfortunately this will also change the poly flow if the edge-loops are not really well defined. This would make a good feature request I think. $\endgroup$ – MarcClintDion Nov 21 '14 at 23:31
  • 1
    $\begingroup$ To find non planar faces you can use the 3D print toolbox addon. Read blenderartists.org/forum/showthread.php?334922-Non-Planar-Quads $\endgroup$ – cegaton Feb 1 '17 at 4:05
2
$\begingroup$

Judging from the normals, they don't fit the visible mesh. See the ones in the red circle they should point more upward (+z). Besides that there should only be one face normal in the center of the face. The area in the orange circle shows two normals but only one face. You should check for duplicated vertices and faces.

Ctrl-V Remove doubles (note the merge distance in the last operator panel)

Ctrl-N Recalculate Normals

enter image description here

Edit after .blend was provided for detailed analysis:

Coloring one of the suspicious faces:

enter image description here

In the wireframe mode you can clearly see that all of the vertices are not on a plane (non-coplanar) this causes the artifacts.

enter image description here

You could try to fix this by selecting these faces and switch to Transformation Orientation to Normal: enter image description here

And scale them with S-Z-Z-0 better you could find out what had caused this.

$\endgroup$
  • $\begingroup$ @Paul in this case select the faces and use 'Flip Normals' $\endgroup$ – stacker Nov 21 '14 at 16:58
  • $\begingroup$ I removed 12 double vertices. If I recalculate the normals, I end up with some pointing inside the mesh, some outside. My mesh has one triangle (two if I apply the mirror modifier), if I remove it and recalculate the normals, they all point the same direction, inwards. The artifacts are still there, even if I flip them. I'll upload the blend file as soon as I get home. $\endgroup$ – Paul Nov 21 '14 at 16:59
  • $\begingroup$ Here's the blend file: pasteall.org/blend/32986 Still unfinished, obviously. The triangle is on the outermost tentacle. I added it because I ended up with a single spiraling loop on the tentacle, instead of rings. I don't know if that would have been ok. Other than that, I was very careful about triangles and n-gons, I always double checked for them, I didn't fin any. $\endgroup$ – Paul Nov 21 '14 at 18:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.