I have trained both SDXL and PonyXL with 800 images using danbooru and BLIP prompt tags.

The zip file contains prompt text file, test images and lora files.

Pony gives the best result

score_9, score_8_up, score_7_up, BREAK

STYLE_FantasticPlanet_ownwaifu, a blue alien with red eyes and a body painted blue with gold chains and a red eye,

(portrait, head tilt, looking at viewer), white border, bokeh, <lora:PONYXL_STYLE_FantasticPlanet_ownwaifu:0.8>, solo,