Type | |
Stats | 1,560 |
Reviews | (168) |
Published | Mar 20, 2024 |
Base Model | |
Hash | AutoV2 13CC3B797F |
PonyDiffusionV6 is nice model but it has little problem on stability of style.
so I finetuned it to anime-like model being stability by ESD.
it named by E(sd)-PONY(diffusionv6)-M(ethod/odel).
so, alias is Eponym.
Difference on other derived models
derived models is stable in style but forgot in some situation (or acts? pose? details?
original model has problem in output too much hands and arms or missing these.
however, this model has no such problem or reduced.
this model is just "finetuned" model. it means style is not perfectly stable.
so you can use with other models as refiner, style LoRA or style tags.
Prompts Rules
basically, same as PonyDiffusionV6.
however, I will add new rules. it's not official but I found it while finetuning.
and also, I don't recommend you over-inputting in negative.
you can reference Examples.
Special_tags
score_tags should use only score_9 in positives,
source_tags should use only in negative,
rating_tags same as PonyDiffusionV6.
Recommend_prompts
positive: score_9, game_cg, anime
negative: source_pony, source_cartoon, source_furry, monochrome, realistic, 3d, cg
Examples
positive: score_9, game cg, 1girl, read examples in description, rating_xxx
negative: source_pony, source_cartoon, source_furry, monochrome