Stable Diffusion 1.5

Stable Diffusion 1.5 is an AI image generation model for artists and developers who want to create artwork or explore design ideas on their own hardware. It can start with a text prompt or modify an existing picture.

• Text-to-image generation: Create images from written descriptions, including photographic styles and artwork.
• Image editing: Use text to guide changes to a picture or turn a rough sketch into detailed artwork, with control over how much the result departs from the original.
• Upscaling: Increase the resolution of images through the image-to-image workflow.
• Local app compatibility: Use the model with ComfyUI, AUTOMATIC1111, SD.Next or InvokeAI, or integrate it into software through Hugging Face Diffusers.

The weights are available in safetensors format for local use. The reference implementation requires a GPU with at least 10GB of VRAM; that requirement applies to the reference code rather than every compatible app. The weights use the CreativeML OpenRAIL-M license, which includes restrictions on permitted uses.

The model works better with English prompts than other languages. It struggles with readable text, faces and precise arrangements of objects, and its outputs can reproduce biases in the training data. The reference generation code includes a safety checker and an invisible watermark. The model card describes research as its intended use and calls for additional safety measures for product use.

This listing is the fixed SD 1.5 checkpoint. Its available model-card mirror identifies CreativeML OpenRAIL-M and explains that it mirrors the deprecated Runway download. The mirror is not affiliated with Runway; the old reference codebase is not evidence that the downloadable checkpoint needs continuing feature updates.