Replies: 6 comments 7 replies
|
How well does this perform compared to stable diffusion? |
|
@comfyanonymous It seems to be a product of russian bank. The Russia is doing a genocidal war against Ukraine, killing, torturing and raping people with a one reason: just being not Russian. It's a country blackmailing a world with nuclear weapons. |
|
@comfyanonymous any update on whether this model will see support on ComfyUI? |
Yandex and mail.ru supporting crimes against Ukraine in all ways they can do inside of Russia. They will remove any hatefull content towards putin / war / etc. They will help to spread misinformation immediately if a random guy from government tell them so. I have used it, so I know how this "propaganda ecosystem" works. You have very bad examples of companies, I would say. |
|
Интересно, что Kandinsky хорошо работает не только с английскими, но и с русскими описаниями. На практике результат сильно зависит от формулировки запроса и самой модели. Я для сравнения разных вариантов иногда использую https://ranvik.ru/image — удобно быстро проверить несколько идей и выбрать наиболее удачный результат. |

Uh oh!
There was an error while loading. Please reload this page.
Ability to write Prompt in more than 100 languages.
Kandinsky 2.0
https://github.com/ai-forever/Kandinsky-2.0
https://huggingface.co/sberbank-ai/Kandinsky_2.0
https://fusionbrain.ai/diffusion
Model architecture:
It is a latent diffusion model with two multilingual text encoders:
mCLIP-XLMR 560M parameters
mT5-encoder-small 146M parameters
These encoders and multilingual training datasets unveil the real multilingual text-to-image generation experience!
Kandinsky 2.0 was trained on a large 1B multilingual set, including samples that we used to train Kandinsky.
In terms of diffusion architecture Kandinsky 2.0 implements UNet with 1.2B parameters.
Kandinsky 2.0 architecture overview:

All reactions