Trained at native 1024px without downscaling on the same dataset as my Pony Diffusion V6 Art Style - "AAM XL Screencap" Lora. Use anime screencap anywhere in your prompt to trigger (this is not optional, this Lora will not do much of anything without the trigger unless you just happen to hit some other caption that existed on an individual image in the dataset).

Recommended strength is around 0.7 - 0.8, but feel free to experiment with going higher or lower.

This Lora (like all my SD 1.5 Loras) was trained against base SD 1.5 and so should be perfectly compatible with basically any checkpoint, regardless of whether or not said checkpoint was photorealistic or anime-focused to begin with.

All showcase images were generated directly at the resolution they are displayed, without any upscaling.