A merge of some of my favorite models and some loras along with an anime model to add lots of variety using danbooru tags
Added workflows to the files as training data
Works well with Euler A at 28 steps
A good resolution is 1364 x 1024
V4 is great, V1 is also great, V3 is pretty bad wouldn't recommend it
V4 Uses a base merge of
https://huggingface.co/Xeno443/3wolfMondAI-SDG
https://civarchive.com/models/827184/wai-nsfw-illustrious-sdxl
https://civarchive.com/models/715287/nova-3dcg-xl
Using the workflow from https://civarchive.com/models/981786/catwai-merge to make it work as eps. 3wolfMond adds more variety/tags for furry stuff
Then ontop of these I have merged
https://civarchive.com/models/974693/realism-illustrious-by-stable-yogi
https://civarchive.com/models/1412827/illustrious-realism-by-klaabu
Along with loras from the previous version
Merge settings: https://imgur.com/a/S2zG2sM
Models used:
https://civarchive.com/models/1412827/illustrious-realism-by-klaabu
https://civarchive.com/models/827184/wai-nsfw-illustrious-sdxl
https://civarchive.com/models/715287/nova-3dcg-xl
https://civarchive.com/models/1257570
Loras:
https://civitai.green/models/1304531/realism-lora-by-stable-yogi-illustrious
https://civarchive.com/models/1638441/lut-color-grading-slider
https://civarchive.com/models/1486904/illustrious-realism-slider
https://civarchive.com/models/1486921/real-skin-slider
https://civarchive.com/models/1333749/add-detail-slider
https://civarchive.com/models/1143824/golden-hour-slider
https://civarchive.com/models/1242944/nostalgic-core-illustrious-pony-and-sdxl
https://civarchive.com/models/1410317/amateur-style-slider
Description
V3 Has worse anatomy. Fingers are more likely to be a weird number, Faces are more likely to be weird. It works well for anthro furry stuff though
FAQ
Comments (3)
IMHO v3 the best, old photo style.
love it
lol really? I found it really bad ay getting 3 fingers or dodgy eyes etc that is the style I like though old photo is what I was aiming for
I for the life of me cannot get this model to produce realistic outputs. They're always some sort of 2.5D or 2D images. Other than that, the model has pretty good output quality.