Kandinsky 3.0 Technical Report

TL;DR

Kandinsky 3.0 is a large-scale, high-quality text-to-image diffusion model with extensions like super resolution, inpainting, and faster inference, demonstrating improved realism and domain-specific performance.

Contribution

This report introduces Kandinsky 3.0, a new diffusion-based text-to-image model with novel training techniques, extensions, and a faster distilled version, advancing image quality and usability.

Findings

Kandinsky 3.0 outperforms previous models in image quality and text understanding.

Kandinsky 3.1 achieves 20x faster inference with no quality loss.

The model supports diverse applications including inpainting and image-to-video generation.

Abstract

We present Kandinsky 3.0, a large-scale text-to-image generation model based on latent diffusion, continuing the series of text-to-image Kandinsky models and reflecting our progress to achieve higher quality and realism of image generation. In this report we describe the architecture of the model, the data collection procedure, the training technique, and the production system for user interaction. We focus on the key components that, as we have identified as a result of a large number of experiments, had the most significant impact on improving the quality of our model compared to the others. We also describe extensions and applications of our model, including super resolution, inpainting, image editing, image-to-video generation, and a distilled version of Kandinsky 3.0 - Kandinsky 3.1, which does inference in 4 steps of the reverse process and 20 times faster without visual quality decrease. By side-by-side human preferences comparison, Kandinsky becomes better in text understanding and works better on specific domains. The code is available at https://github.com/ai-forever/Kandinsky-3