Update: V3 released.
Get Animated!
Aijalon Anitoon is a flexible style model based on an interpretation of my own drawing style, or how it would look if I didn't suck. It was trained with an amalgam of months' worth of my own scribbly drawings slowly refined at low CFGs in img2img, in conjunction with a prompt cocktail recipe I spent countless nights experimenting with to get the closest approximation possible of the picture in my head that my unsteady hands and horrendous line control rarely live up to.
The LORA is extremely flexible and useable with other SDXL and Illustrious LORAs. It outputs respectable results with and without the standard (masterpiece, best quality, very aesthetic, absurdres tags), and most solo focused outputs look good without the need to use a hi-rez fix.
Anitoon is highly scalable, and the output strength can safely go from :0.1 near standard base model with subtle tweaks, to a terminally puni plush :1.7 before outputs start abstracting from the sheer disgustingly cute moe energy!
The output can be further tweaked and refined with the inclusion of the trigger word "aijalon" somewhere in the prompt.
The closer to the beginning of the prompt the more powerful I become!
Recommended CFG: 5.5
Recommended Sampling Steps: 25-30
I know absolutely nothing will stop people from using this LORA however they want so go nuts!
All I ask is to be credited if you commercialize outputs created using this LORA.
Description
Aijalon just got more Anitoon...er!
Trained on 999 hand edited images!
Fewer artifacts
More detailed eyes
more accurate full body angle details
Fewer weird half rendered characters poking out of the background
Better male characters
Wider variety of body types
FAQ
Comments (7)
Works great! Makes grimdark bloody gore cute again. It can overpower some other Loras but still works well with them. I got the best results by adding it to the least-detailed prompts.
I had a lot of fun making images with this LoRA. Its a refreshing breeze of cuteness. Definitely will be using this more!
v2 of this LoRA taught me that it's possible to improve on perfection and the result is even better perfection. Incredible job, Aij.
It's really fun to use this LoRA, I feel like I'm going to use it a lot, great work friend.
I love this model, the output is super cute
using this model feels like cheating, I can throw it on at like 25% and use the worst generation settings and get great results
This is amazing. The idea of refining your own drawing and then use them as a dataset is marvelous 💖💕