Trained with 12 images, using just the standalone keyword is sometimes enough, but using more specific descriptions yields better results.

Use less weight for custom outfits.

koyorinatsu, eyewear on head, see-through jacket,