Stable Difusion

  • latent-difusion model - The main diference of this model with respect to the other ones is the use of a latent difusion model and that it performs image modification as it can perform operations in its latent space
  • Stable Difusion consists of two parts: the text encoder and the image generator - The image information creator works completely in the latent space
  • This property makes it faster than previous difusion models that worked in a pixel space