The last few years can be described as the heyday of generative models working in different modalities. This report is dedicated to one of the most
We’ll discuss the theoretical aspects of the diffusion process, its advantages and disadvantages. We will also talk about the new architecture of Kandinsky 3.0, its training peculiarities, the nuances of collecting, filtering, and storing training data, and the results obtained. We will address the issue of generating videos from text descriptions, discussing creating different types of animations and the