This is the workflow for Ratatoskr Flux it also include the Custom clip_l of the model that i highly recommend because its from the original Ratatoskr SDXL model trained partly on the same dataset.
You can test the workflow and model here:
Ratatoskr - Creature - Generator [Flux Version]
please check out my other works and if you want sopport me give a like ;-)
Description
V1 of the workflow
FAQ
Comments (5)
You can test the workflow and model here:
truly your Flux model is a bit darker then the base model of Flux and nearly don't need to refine at only one person in a portait, at 1080p Ratatoskr Flux looks like an image from my smartphone
And Flux is at me now near the same fast then normal SDXL without turbo or lightning, so around 1-1,8 sec/it comfyUI is in the latest version 2630 really good for Flux.1
How does one install this? A brief explainer in the description would be helpful to people who are not as experienced with comfyUI.
This is actually a good idea:
So first of course install comfy
Then download the comfy Manager
Open the json file in comfy go in the manager and push search missing notes. Restart comfy. The checkpoint needs to be in comfy/ model/ checkpoint .
Also the sdxl model
For the clip_l and t5xxl put them into / models/ clip
Open comfy again select the flux model
The sdxl model
Load both clip in dual loader.
So now it's ready to run.
note: loading 2 models need at least 32 or more ram and at least 19 GB vram .