Anime Style Backgrounds for Pony Diffusion

This experimental LoRA is intended to generate detailed anime style backgrounds with Pony Diffusion checkpoints.

Can be used generate background art, and to reduce the occurance of simple or white backgrounds when prompting for character subjects.

(simple_background, white_background are common in booru training data for characters)

New in v2:

This version is trained on AutismMix SDXL

Aesthetetic quality modifier tagging has now been applied to the training data, in addition to the regular Pony score modifiers, include at the end of your prompt:

masterpiece, best quality, very aesthetic, absurdres

*This is part of an ongoing experiment to finetune Pony diffusion towards quality tags that allow it to be merged smoothly with other SDXL models.

Prompt structure

Positive prompt (note quality modifiers at end of prompt):

score_9, score_8_up, score_7_up, score_6_up, source anime, scenery,
{{natural language description}}
{{tags}}
masterpiece, best quality, very aesthetic, absurdres

Negative prompt:

score_4, score_5, score_6, monochrome, censor, 3d, render, CGI, text, watermark, cartoon, sketch, painting, impasto,

Trained on AutismMix SDXL

Put the quality keywords at the end of your prompt:

masterpiece, best quality, very aesthetic, absurdres