Generate images with the original Stable Diffusion 1.5 for free, without an account. It produces 512×512 pictures in a few seconds, reacts strongly to keyword prompts and artist names, and remains the most customised model ever released. Useful to compare with SDXL and Stable Diffusion 3 on the same prompt.
SD 1.5 was trained at 512 px. Larger sizes produce duplicated subjects; upscale the result with the upscaling tool instead.
Yes, the members page runs Stable Diffusion 1.5 on our own GPU with seed, size and steps control and a private queue.
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: What is Stable Diffusion?