Kandinsky: an Improved Text-to-Image Synthesis with Image Prior and Latent Diffusion

TL;DR

Kandinsky1 is a new latent diffusion-based text-to-image model that combines image prior techniques with a modified autoencoder, achieving state-of-the-art quality and supporting diverse generation modes.

Contribution

The paper introduces Kandinsky1, integrating image prior models with latent diffusion and a modified MoVQ autoencoder, along with a user-friendly demo and open-source release.

Findings

Achieved an FID score of 8.03 on COCO-30K dataset.

Demonstrated high-quality image generation with diverse modes.

Outperformed existing open-source models in quality.

Abstract

Text-to-image generation is a significant domain in modern computer vision and has achieved substantial improvements through the evolution of generative architectures. Among these, there are diffusion-based models that have demonstrated essential quality enhancements. These models are generally split into two categories: pixel-level and latent-level approaches. We present Kandinsky1, a novel exploration of latent diffusion architecture, combining the principles of the image prior models with latent diffusion techniques. The image prior model is trained separately to map text embeddings to image embeddings of CLIP. Another distinct feature of the proposed model is the modified MoVQ implementation, which serves as the image autoencoder component. Overall, the designed model contains 3.3B parameters. We also deployed a user-friendly demo system that supports diverse generative modes such as text-to-image generation, image fusion, text and image fusion, image variations generation, and text-guided inpainting/outpainting. Additionally, we released the source code and checkpoints for the Kandinsky models. Experimental evaluations demonstrate a FID score of 8.03 on the COCO-30K dataset, marking our model as the top open-source performer in terms of measurable image generation quality.