The latent diffusion paper was co-authored by Patrick Esser while he was a researcher at Runway

General research and information gathering from multiple sources

Generative model learning to reverse a noise-adding process to generate high-quality samples.