BunnyMint is a Pony model based on two finetunes. The first finetune was trained on 8k randomly selected artist images with score:>99 from danbooru for 10 epochs to improve overall aesthetic. The second finetune was trained again on 20k hand picked NAIv3 images for 20 epochs. Personally trained Blue Archive LoRA was then merged on top to make style more alligned with personal preferences.

HF: https://huggingface.co/lylogummy/BunnyMint

Donate

Finetuning models on personal hardware is pretty costly, so if anyone likes what I do and feels like it, here's my gumroad:

https://lylogummy.gumroad.com/l/avhut

Training

(8k random artist images score:>99) * 10 Epochs

(20k NAIv3 dataset) * 20 Epochs

Total steps: ~480k

Usage

Sample prompt

score_9, score_8_up, score_7_up, source_anime,
1girl, BREAK

joyful, black skirt, white pantyhose, crop top, double v,inside submarine, legs apart, pleated skirt, jumpsuit,
masterpiece,best quality

Negative prompt: 
nsfw,censor,extra_fingers,logo,score_4,score_5,score_6,lowres,(bad),text,error,bad hands,fewer,extra,missing,worst quality,jpeg artifacts,low quality,watermark,unfinished,displeasing,oldest,early,chromatic aberration,signature,extra digits,artistic error,username,scan,[abstract]

Settings

Steps: 40 
Sampler: Euler a 
Schedule type: Polyexponential 
CFG scale: 6
Size: 624x912
Hires upscale: 1.5

Since the model is Pony-based, anything Pony related should work here as well.

License

This model is licensed under "Fair-AI public license 1.0-SD", please refer to the original License for more information: https://freedevproject.org/faipl-1.0-sd/