because SDXL sucks at fundoshi

1.5 images sample images upsampled.

The 1.5 model is preferrred

SDXL INSTRUCTIONS

• Likes turning into a loincloth style fundoshi below 0.7 LoRa strength (working on a fix)

• Try LoRa strength 0.6 - 0.9 and cfg 4 - 30 with 0.7 and 10 as your starting point

• The suggested resources are not necessary and were used to fix some issues with hands

the cover's data is trying to find the prompt for two images, the realistic one was prompted by using words like "high definition, realism, realistic"

https://ko-fi.com/alightbulb if you're feeling generous