Zem Priestess
My first serious attempt at Stable Diffusion rendering pipeline. Found this old model sitting on my HDD collecting dust and decided to see how SD could be applied for visdev needs.
Super fun process, although very tedious one. Literally hours spent on little increments and picking the right option. Not easy! Need a better hardware.
Most of this done in SD with leonardo.ai used as last 5% beauty filter over SD rendition.
Awesome tool, awesome workflow - I haven't had so much fun playing with my art in a while.
Cheers.
final color corrected art
unprocessed pic from SD
Models used: Dreamshaper, URPM
Controlnet with normalmap and lineart inputs. Upscaled and detailed using inpaint.
Source image - combined Marmoset raytraced rende
Normal map pass, adjusted for SD
lineart generated from Normals with fakeCrazyBump photoshop script
Final art with more stylized render, staying closer to original design
Zbrush BPR render
Leonardo.ai pass
Zbrush render
leonardo.ai
leonardo.ai
leonardo.ai
Stable diffusion URPM model with Dreamshaper combo
URPM model
Source Zbrush BPR render
Found that SD works better with stylized renders like this, giving itself more freedom to interpret the light.
Lineart generated in Zbrush with outline filters
much better looking depth pass from Zbrush