Kandinsky 2.2

Kandinsky 2.2 is an open-source and multilingual latent diffusion model. An upgrade to Kandinsky v2.1. It is created by a group of researchers from Russia. It’s not based on Stable Diffusion, it’s an original model. It has a base resolution of 1024x1024 pixels. It’s significantly better than Kandinsky v2.1 on realistic images, images with small detail and images that depict wider scenes. It can handle different aspect ratios and resolutions very well without repeating patterns.

Kandinsky 2.2 can be considered an open-source Midjourney alternative, just like the previous version. However 2.2 is more inline with Midjourney v5 because of the improved realism.

Kandinsky 2.2 Examples#

Here are some images created with Kandinsky 2.2 on Stablecog.

Try Kandinsky 2.2#

You can try Kandinsky 2.2 on Stablecog for free. Just click the button below.

Kandinsky 2.2 on Stablecog Gallery#

You can also see more examples of images created with Kandinsky 2.2 in our gallery by clicking the button below.