How to Generate Images Using Diffusion Models

There's been a work done on a diffusion models, and i think s came out of gogbrain a. Andis kind of a new architecture for synthesizing images. The way that thi works is you start with a completely noisy image. You train the motel to remove this noise and ow, you can caiv do this in little steps,. Eventually, you know, you can start with a fully noisy image and then create y san  the way you describe it. It does feel wrong in a certain sense, like a human step.