Kandinsky-2 icon indicating copy to clipboard operation
Kandinsky-2 copied to clipboard

Kandinsky 2 — multilingual text2image latent diffusion model

Results 82 Kandinsky-2 issues
Sort by recently updated
recently updated
newest added

Hi, I have a problem with threading the model generate similar images. where does the race condition happen? i for now I know it happened when calling generate_img function in...

Есть ли возможность делать генерации с вручную вбитым зерном?

The title pretty much speaks for itself. Calling `.generate_inpainting(...)` with `batch_size` higher than `1` throws an error. Log: ``` [/usr/local/lib/python3.9/dist-packages/kandinsky2/model/text2im_model2_1.py](https://localhost:8080/#) in forward(self, x, timesteps, inpaint_image, inpaint_mask, **kwargs) 150 inpaint_mask =...

![image](https://user-images.githubusercontent.com/4674698/229825080-bcd92583-6798-442f-b799-9512d7279b3e.png) As you can see the init does not accept img2img despite it being supposedly supported.

При запуске пример ```python from kandinsky2 import get_kandinsky2 model = get_kandinsky2('cuda', task_type='text2img', model_version='2.1', use_flash_attention=False) images = model.generate_text2img( "red cat, 4k photo", num_steps=100, batch_size=1, guidance_scale=4, h=768, w=768, sampler='p_sampler', prior_cf_scale=4, prior_steps="5" )...

hi guys, first of all great work! two questions 1. can I safely say Kandinsky works best with Russian, meanwhile also working with English and potentially other languages? i.e. Russian...

Are there any plans to add Animation and/or initial images? Would be a huge asset. And thanks for your excellent work.

I contrasted some of the differences(and Midjourney) (1) .about person face ![image](https://user-images.githubusercontent.com/48466610/230591139-d3641c8a-070a-4ebe-b8e7-add0ed919971.png) ![image](https://user-images.githubusercontent.com/48466610/230591280-6446f94a-ff2b-4210-bd9e-3ef405b72de8.png) (2) They are very similar in terms of scenery