Trained on 400 images of the legendary Ninjartist. I've always been a big fan (and supporter) of his work, so it's been a goal of mine to try and mimic his style in AI. Of course it'll never be good as the OG, but it's fun to play with!

The Lora works well enough on it's own, but it a bit hit and miss when mixed with other Loras. I'm already working on a V2 that has been more carefully created to hopefully solve that issue.

I've added a with/without lora comparison in the images to show how the lora at full strength affects the prompts. I think it does well at emulating the soft squishy breasts, innie pussies and gorgeous lighting of Ninjartist's style.

Recommended parameters:

Checkpoint: Pony V6

Sampling steps: 20-30

CFG: 4-7

Lora weight: 1

Positive prompt: score_9, score_8_up, score_8_up, detailed illustration, realistic

(optional: zPDXLxxx, zPDXL2, OverallDetailXL)

Negative: score_6, score_5, score_4, pony, bad anatomy, bad quality, lowres, badly drawn eyes, badly drawn hands, badly drawn feet, extra limbs, missing limbs, extra toes, extra fingers, missing digits, chibi, teeth, monochrome, camera frame, text, watermark, signature, multiple frames, 3d

(optional: zPDXL2-neg)