Generate 1024×1024 images with Stable Diffusion XL 1.0 for free, directly in this page and without an account. Type a prompt, optionally a negative prompt, and get a picture in about 15 to 40 seconds. SDXL is the model that reacts best to artist names and detailed style words.
No hard limit, but the demo shares a GPU with other visitors: at busy times you wait in a queue. Members get a dedicated queue on the txt2img XL page.
SDXL is trained at 1024×1024 and works well at any 1-megapixel aspect ratio such as 1152×896 or 896×1152. Smaller sizes give worse compositions.
This demo is hosted by a third party (Hugging Face Space) and embedded here; it can be slow, queued or temporarily unavailable when the host is busy. Nothing to install, no account needed.
Read the guide: SDXL vs Stable Diffusion 3 vs SD 1.5: which model to use?