Train a Lora style model using 30 anime light and shadow style images at a 1024px resolution, repeat5, bs6, and 5e-4 with 109 epochs. No prompt words are needed during use; set cfg to 1 and adjust the model weight to 0.5~1. It is recommended to use a larger output resolution and apply a 0.55 redraw magnitude with 2x upscale using hires fix (similar to my cover image).

AnimeLight-v2

Train a Lora style model using 30 anime light and shadow style images at a 512px resolution, repeat5, bs6, and 5e-4 with 109 epochs. No prompt words are needed during use; set cfg to 1 and adjust the model weight to 1 (you may experiment with other weights as well). It is recommended to use a larger output resolution and apply a 0.55 redraw magnitude with 2x upscale using hires fix (similar to my cover image).

Files with SHA256: 19ba9520f6a54b2e9ce77633cf6be42da5f51ad3085465d945bd9f3f0e41d087