Stable Diffusion v1-4 (Original)

Stable Diffusion v1-4 Original is an open-weight latent diffusion model from the CompVis research group that generates images from text prompts. It is the canonical CompVis checkpoint, fine-tuned from v1-2 for 225k steps at 512×512 resolution on LAION-aesthetic data with classifier-free guidance. Source: supabase.