Home Artists Posts Import Register

Content

I am in the works of testing out Stable Diffusion XL for my NSFW works,

You may have heard about it in my anniversary post about my Psychedelic art being made on it.

SDXL is a majorly improved version of Stable Diffusion 1.5 where it's max image size is 1024 x 1024, while Stable Diffusion 1.5 is 512 x 512 and 768 x 768 in size.
512 + 512 = 1024 means you can pick an image ratio up to the total amount to 1024 and same for the 768 x 768, and also SDXL's size.

It also has much more accuracy on the overall design/faces/hands/body/anything really of what you want. Also has a much better understanding of words which means it can take in more natural language than SD .15 & SD 2.0(far less superior than 1.5) and even more prompts at once. This image clearly shows the feet looking very good and fingers over all looks great too!


SDXL is a 6.36 GB model while SD 1.5 is usually 2GB in size nowadays, (back then it was the same size of SDXL but not for long.


SDXL takes up more performance over the other models but generally more worth it to use if you want really good results. Forge A1111 has given me WAY better performance on a 8GB VRAM 3070 GPU so doing this is much more feasible!

I plan to test out other things and see if I can get characters in sex positions as well.

Stay Tuned!

Files

Comments

piales

Good luck ❤️‍🔥

Lanakila

How are you getting around the resolution change, having to retrain all your loras for use in XL?

Rule34 Diffusion

Sadly using loras for 1.5 on XL doesn't work, it requires a new lora made from scratch that's trained on an SDXL model in order for it to work.