SDXL / XL Models
The "XL" models on PixAI — bigger images, better anatomy, better prompt understanding.
XL models on PixAI are built on SDXL, the upgraded generation of Stable-Diffusion. If you've been frustrated by murky details, wonky fingers, or the AI not understanding complex ideas like "holding a pair of chopsticks" — XL models are the fix.
What you get
Larger images at higher resolution
Noticeably better human anatomy (fingers included)
Better understanding of complex compositions and poses
What to watch out for
Higher credit cost. XL models can't generate small sizes — they start at 768×1280 pixels, so each image costs more Credits.
XL LoRAs only. XL models only work with XL LoRAs; regular LoRAs won't attach. Can't find the XL LoRA you want? Train one — see LoRA-Training.
Some features don't apply. A selection of ControlNet methods, Composition, and Outpainting aren't compatible with XL models.
Prompts behave differently. Your trusty SD 1.5 prompt may need rework. Read the model's description and learn from prompts other creators used with that model.
Tip
Like an XL model's accuracy but not its aesthetic? Generate a base image with XL, then use it as a reference image with your favorite regular model — see Img2img.