[ATTACH=CONFIG]486952[/ATTACH]
Confirmed 2 things. Normal map plugged into diffuse post SSS creates shadow artifacts. Normal maps also seem to greatly exaggerate render time. Computation of scale? Something to do with how SSS handles normals? Render result without normals looks good. Maybe I don’t need them…on to the next test.
Various tests for changing normal map node…none satisfactory.
[ATTACH=CONFIG]486958[/ATTACH]
[ATTACH=CONFIG]486960[/ATTACH]
Unable to illustrate point with pictures at this time due to “server settings”. A shame. Good for research to have cohesive place to review notes. My desktop is cluttered with fiance’s cat videos and various things that apply to her profession.
Maybe make a blogger account…
In any case, further testing has led me to believe that the normal map itself is the issue. Adding in standard displacement caused “waves” on subject’s forehead and cheek. Unacceptable.
2nd test: Removed standard displacement map. Added “micro” displacement map. Tiling an issue. Shadow artifacts returned. Unacceptable.
Conclusion: Test 217a and b reveal that the problem is likely with mapping the normals to the mesh itself. Bump map was generated, but normal map was created by me. Same with the “micro” displacement map. The idea is to capture skin imperfections (the “weave” of skin) and pores up close. Created a blown up image of skin. Used Gimp to create normal and displacement maps. Maps were not to scale of mesh. Correcting scale with nodes, I believe, is causing a computationally complex situation for Cycles. Hence the early issues I encountered and didn’t understand.
This likely means starting over, as previous test took less render time, but had issues with the mesh deforming oddly.
Next step: re-map normals and micro displacement to full mesh.
Reviewing rendered normal node, normals appear to be more of contour lines than what I had previously realized. They are the true issue as the “lines” of the normals are flowing down the forehead in a way that makes it appear that I have tiled images. I may just need to rework the normal map. Attempting a render without it.
Render proved inaccuracy of observation. Tiled images of normal map + displacement map make flaws obvious when subserf modifier is used.
The problem continues to be that the images are tiled and not mapped.
Test 225:
I believe I have managed a work around for having to build another normal/micro DP map(s). If I use the current maps and mix them carefully with maps generated from the diffuse (no fine details) I should be able to slightly layer them into the skin. [ATTACH=CONFIG]486970[/ATTACH]
“Corrected” displacement map. Not satisfied with this result alone. Will need to test results with different combinations of normal settings.
OF NOTE: when I’m right, I’m right. Pixelated shadows are gone. Perhaps shadows for low poly meshes are limited in detail.
Also of note: this image is without subserf modifier, low samples and about 10 seconds to render.
Performance boost for reducing, not removing, micro maps. Interesting.
226:
“Corrected” Normals and “Corrected” disp.
[ATTACH=CONFIG]486971[/ATTACH]
Good proof of concept. Changed lighting to show more of face. Shadows fine. Needed to see how the material was applied to mesh. Some lines okay. These lines…not so much. Can lower intensity of normal map.
FFS it looks like she has a 5 o’clock shadow (moustache)
Previously mentioned issue was stupid. Lines from Disp. map noted previously. Thought they would go away or blend with normal map. This did not occur. Spent much time trying to adjust normals only to remember it was disp map issue. Lowered strength.
Made several adjustments to HM, NM and DISP. maps.
SSS not playing well with disp. I think. Awaiting final render for further analysis.
Test 2 a wash. Technically, test 3 is a variation of 2.
Result would be acceptable if this was 2001 and I was making a video game.
Skin has a plastic appearance despite blended normal map.
Considering re-working normals (again).
New test has cleaner workflow and more experience understanding the purpose of each node.
I’m told that this will get easier the more I experiment.
Time will tell.
Test 3.
Reset nodes. Trying something new. 3 sss layers. Diffuse map. Normals after rework.
Already optimistic results. Not sure how normals are going to look added into this, but…
Note variation in this last image vs. final image from last test. Same lighting. Same scene. Same mesh.
Brightening the skin wasn’t something I was aware I needed to do. By comparison, I can tell you that I did.
Edit: I wanted to go ahead and add, that having a “map” for each layer of SSS did little in regard to quality. I made these maps and put a great deal of effort into their creation and design, colorizing high points on her body, not just her face, only to come to this test where I removed the maps and simply changed the color of the SSS layer on the node. It is much simpler and, with fewer maps to load, quicker to render, but frustrating all the same. I’m guessing that the maps are unnecessary when the SSS nodes are mixed with the diffuse.
Had I been working on this regularly, it might have taken a good deal of time to realize this.
With micro-geometry. Normal maps had issues with previous test. Here they are seamless. Will experiment with proper placement. Also of note, geometry was negatively impacted when normals and displacement maps were used in previous test.
Here, there’s little difference between this render and the last. Next test is to put normals further down the pipeline.
Mixed feelings about this result. Plastic feel is gone. Skin has a more realistic feel…but still feels off.
Moved normal map to Glossy layer.
Going to try adding them somewhere else, too, but need to avoid a possible overpowering.
Surprisingly successful result adding properly scaled normal map to diffuse.
Still using micro geometry on the glossy layer and displacement.
Was pleased. So pleased, forgot to save rendered result.
Immediately went to light testing.
So…here is that result with Full Global light paths.
Going to test with primary colored lights and then test same lights from a profile view to ensure quality. Next step is to try different skin colors. Then fixing those eyes.
Increased micro displacement. Still not 100% with result. May need to remake micro-displacement map. Awaiting HDRI render to verify displeasure.
Displeasure verified.
Notes: displacement OK. Not great. Obviously tone needs to be adjusted. Not an issue. Frustrating that I have to render these out before I can adequately judge results.
I made adjustments to the tone in the image texture because changing the color values in the SSS shaders seemed to have a minimal impact. I also completely reworked the normal map, micro displacement and bump map.
I will have to wait for the drunk to wear off before I can adequately assess this image…but drunk me likes it.
Message for sober me: fewer freckles, man…fewer. freckles.
Moved some nodes around. Removed extraneous nodes. Removed redundant nodes. Changed camera angle because creating maps requires different camera settings and I forgot to copy down the camera’s previous location. Removed microdp map. Used HM in displacement socket.
Oddity: light value unchanged, skin appears darker. Did not modify tone.
Duplicate.
This node setup works, it seems, exclusively for white skin.
Darkening the skin required changing how the nodes deal with specularity.
It’s funny, I thought changing the image texture to a “dark” skinned texture map and changing the tones in the SSS shaders would be enough, but it’s far more involved than that. The “deep” layer SSS I’ve been using to create red tinted areas like lips or ears is useless. There’s just a number of differences that cannot be simply overlooked.
I thought it would be simpler…
Test 4 begins.
