YOU HAVE TO USE IT WITH "b74_aes"

Use with strength of 1.4+ when only using this one lora, reduce to 1 when use with other lora

This lora emphasized on lighting and shadow(anime stlye).

I only tried this on flux dev fp8, not sure how it works on other model.

recommend also using with this lora AnimeLight-v2 - AnimeLightv2 | Stable Diffusion LoRA | Civitai

Will try train again with 1024*1024 and more steps if I could get more buzz.

I picked 450 my most favourite aesthetical anime imgs, feel free to ask me for the dataset and train it yourselves.

my discord: b74

here is the other epoch if you want: aesthetical_epoch_1.safetensors · b74444/style1 at main (huggingface.co)

aesthetical_epoch_2.safetensors · b74444/style1 at main (huggingface.co)

aesthetical_epoch_4.safetensors · b74444/style1 at main (huggingface.co)