If you are a true artist please try the following settings:
-Sampler: Seed3 (almost NEW in comfyUI)
-Scheduler: Beta ( I know its experimental...or maybe not)
-Denoise: 0.75
-CFG: 1.5-1.9 (Trust me! And Trust the process!)
-Steps: 10-14 ( It's enough, because Seed3 form the latent in the first steps of iteration)
-Enjoy high quality!
Recommended LORA-s for this Checkpoint
-https://civarchive.com/models/768917?modelVersionId=1278054
-https://civarchive.com/models/2140417?modelVersionId=2421065
-https://civarchive.com/models/1400090?modelVersionId=2524593
-https://civarchive.com/models/1429923?modelVersionId=1666165
All LORA weight max 0.30, because Seed3 strenghten LORAs too!
This model is not a single person’s creation, and it is not solely my own work.
It is a fusion of the efforts, talent, and dedication of many great artists and model creators.
This model is a merge of:
-Popwind Toon v3.0 by ERYTH_AI
-Tulpa Toons v3.0 by ViciousTulpa
-Dream Brush v2 by King_Hatchet
My goal was to create a mash-up that combines cartoony aesthetics, solid anatomy, believable weapon handling, and Disney visual quality into a single cohesive model.
The model excels at generating both female and male characters, as well as monsters and fantasy creatures. It follows prompts very well and handles LoRA stacking exceptionally, even at low weights.
Beautiful results can be achieved with low CFG values and low step counts. Its muted color palette gives the images a subtle, almost intangible atmosphere that enhances the final result.
This is my 5th project of this kind.
Please enjoy it.
Description
I want to introduce you, my next version of Crystal clear DisneyxAnime merge. This version is focusing pixar aesthetics, but kepp the lovely vibe of previous versions. Same good prompt adherence, character aesthetics, but in new custome. Living backgrounds, extremely detailed eyes, and perfect anatomy, with great weapon handling. Please enjoy generating with this model & share the results with me.
FAQ
Comments (4)
The image gets messed-up at output after normal generation during the steps. I suppose this is because the VAE isn't baked in into the model, is it? If so, which VAE you suggest for best results?
looking at a workflow in comfyui he has sdxlVAE_sdxl_VAE.safetensors
I just did one with sdxl_vae.safetensors and it worked,
So the sdxl one seems to be fine.
I'm still new to all this so dont roast me on VAE stuff, i am learning.
@pezpok1929 Oh yes ! In description, I sayd VAE is not baked. Use sdxl VAE for best results!
VAEis not baked in this. Use SDXL VAE. Works like charm.