This workflow generates an image using flux and then uses Z-Image-Turbo to enhance features such as skin. I'm expecting this to work much better once the full base model is out as well as with finetunes and loras.
Flux is great at producing a variety of outputs from a single prompt whereas ZiT seems to be kinda same same even across different seeds, so I decided to use them together.
you will need these upscalers:
- https://openmodeldb.info/models/1x-ITF-SkinDiffDDS-v1
- https://openmodeldb.info/models/1x-SkinContrast-High-SuperUltraCompact
- https://openmodeldb.info/models/4x-classicalSR-DIV2K-s64w8-SwinIR-M
I use these luts in particular but you can use whatever is appealing to you
- https://freshluts.com/luts/34
- https://freshluts.com/luts/2729
Description
Much better quality with this version.
1 FLUX sampler and 1 ZiT refining pass. I added the lying sigma node and it seems to help massively with the quality of ZiT.










