Michele LaurelliBlog

Latent diffusion

Latent diffusion: Diffusion performed in a compressed latent space rather than directly on pixels. An autoencoder maps images and latents, reducing generative cost. The decoder reconstructs pixels and compression affects recoverable detail. Examples include Stable Diffusion, Denoising in VAE space, High-resolution generation.

In italiano: Latent DiffusionLegge II · Probabilità, softmax e modelli generativi

An autoencoder maps images and latents, reducing generative cost. The decoder reconstructs pixels and compression affects recoverable detail.

Examples

  • Stable Diffusion
  • Denoising in VAE space
  • High-resolution generation