Created a workflow for ultrarealistic results with Z-Image. While I designed it for my NiceGirls LoRA, it actually works good without people too (like on 1st picture). Actually workflow non-optimized for speed, cause i aimed at quality.
U need to download my lora: https://civarchive.com/models/1862761?modelVersionId=2465980
And custom nodes: https://github.com/ClownsharkBatwing/RES4LYF
https://github.com/PGCRT/CRT-Nodes (not sure if you still need this pack to use z-image lora loader, because as I understand it, in the latest versions of ComfyUI the native LoRA Loader already supports z-image.)
Actually you can try without my lora too
Description
FAQ
Comments (10)
I'm having trouble getting the 'LoraLoaderZImage' node downloaded, the manager isn't catching it either or where to find it nor is it in the custom node library you linked
have the same problem, don't find that node anywhere lol
same here
@srrelo1 @1GirlUniversity here is node pack https://github.com/PGCRT/CRT-Nodes
UPD: Thanks to @theqqnomore733 for the explanation. Apparently, the latest ComfyUI versions natively support Z-Image LoRAs, so you can simply swap LoraLoaderZImage for default LoraLoaderModelOnly
@Danrisi was wondering the same. Thanks!
@Danrisi Is this updated?
Can't get results even close to this, using the ae vae and the rest is the workflow. Tried all LoRa loaders I can find.
very good workflow, thank you
The Flux.1 Dev Resolution node isn't working. It has an invalid value of 2.0 (not selectable in the drop down box). So I bypass it and obviously images are only generated at 1024x1024. Any ideas?
Sorry, what makes this WF special WF suitable for realism? There are just ordinary nodes, as elsewhere