Activation tag is nyantcha, made to work with autismix so it might be a little overcooked with base ponyv6.
If you are interested in making your own loras here are some training settings to use with Kohya GUI:
Tags settings: https://files.catbox.moe/ocsne4.json
No tags settings: https://files.catbox.moe/nglzg5.json
They are not beginner or low VRAM friendly settings. First one requires bitsandbytes 36+ and second one requires bitsandbytes 0.35.
I highly recommend resizing the lora with these settings in Kohya GUI if you use the no tags settings: image
Tagging was done using SwingV2 at 0.35 weight and 0 min tag fraction with nyantcha, as prefix tag. Using 1000-2000 steps is also fine depending on the artist.
Description
FAQ
Comments (10)
Nice style! Please make more!
can you tell me how to generate similar styled images? what are u generating them on? I am using comfy ui and they come out uncanny and hyper-realistic doll like
What is "score_9, score_8_up, score_7_up" in your prompts?
It's a pony diffusion thing, I am not really sure does, but it helps with making a good image. Only works for pony diffusion.
AutismMix_ConfettiChaosDPO might be a derivative of pony diffusion. That's why they are using those tags.
The scores are basically just quality tags. NovelAIv1 introduced the "masterpiece" and "best quality" tags and that's why you see a lot people use those quality tags on most SD1.5 models. So instead of using "best quality, masterpiece" in your prompt box you should just use those score tags. Also, you don't need to put stuff like "low quality, normal quality, worst quality" in your negative box, it doesn't really change anything since PonyXL or AutismMix don't use those quality tags.
And I forgot to mention you actually don't even need to use any negative prompts, which was mentioned in the PonyXL description:
"The model is designed to not need negative prompts in most cases and does not need other quality modifiers like "hd", "masterpiece", etc..."
I recommend reading the full description on the PonyXL model so that you get a better idea of what it's capable of
Did you train this model on AutismPony? I'm asking because whenever I try to train a loRA on autismPony (so that I can use loRA based on it) I keep getting: Error: The model you selected is invalid or corrupted, or couldn't be downloaded
Buen lora
This is somehow still the best Nyantcha/thiccwithaq lora outthere
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.