Model Summary
This anime-style SDXL model merges mine Mistbloom LoRA with the Illustrious-XL base checkpoint. Optimized for Stable Diffusion XL, it works best with Euler sampling and supports further LoRA customization.
Ideal for: Anime
Key Features
✔ SDXL-optimized anime styling (via Illustrious-XL + Mistbloom LoRA merge)
✔ WebUI & ComfyUI compatible
Recommended Settings
📌 CFG Scale: 3–5
📌 Sampler: Euler (best results)
📌 Steps: 20–35
Optimal Resolutions:
1024×1024 (1:1)
1254×836 (3:2)
1365×768 (~16:9)
Merge Details
Base Model: Illustrious-XL
Merged Component: Mistbloom LoRA
License & Restrictions
🚫 Non-Commercial Only
Generated images cannot be sold or used commercially.
⚠ Ethical Use Required
No harmful/illegal content. Users assume full responsibility.
(License terms inherit from Illustrious-XL.)
Description
Fix discolour issue. Thanks Julio912 for the feedback.
FAQ
Comments (8)
I recommend that you bake a VAE so that the color is much more noticeable and is not so monochromatic.
Thanks, just noticed my Monitor settings were messed up, I will update modal
And the white spots are artefacts from the missing VAE as well ^^
After taking a few weeks to go fuck around on text gen, I return to Stable Diffusion, download a bunch of models and low and behold, gotta scratch my brain to make them work. On the other hand, this model works like a god blessed early 2000s Toyota. Requires absolutely zero tinkering, zero fiddling, I literally load the model and hit generate on any prompt and bam clean image.
This model is what I use to test my workflows to see if I fucked up or the model I'm using is just acting up for no reason. God-bless this author and whatever witchcraft he was using to create this model. Shit's a godsend for me o7.
What setup are you using?
auto1111 webu, Comfy UI, Forge, Reforge, or other?
@Rustyvivid auto1111. Thought I wrote it in but ig through the numerous revisions I wrote I kinda forgot about it and left it out.
As far as I know, Auto1111 has stopped updating. I would recommend reForge
This is what I use to test the checkpoint on WebUI, If you have issues with eyes, you can use ADetailer, this option is usually below "Seed "
@Rustyvivid Ye, I've been using adetailer and it dosent seem to fix it. Probably just an auto1111 thing. I really should try another UI but the one I did attempt to switch to had such bad extension support I had to switch back. That being said even 'stock' without artist style loras your model works very well and quick.