Stable Diffusion uses a variational autoencoder (VAE) to generate detailed images from a caption with only a few words. Unlike prior autoencoder-based diffusion models, Stable Diffusion incorporates a ...