Kandinsky-2
Kandinsky-2 copied to clipboard
Kandinsky 2 — multilingual text2image latent diffusion model
I have finetuned prior and unet models for Kandinsky 2.2 inpainting with Lora for about 15 images. The results are bad compared to sd 1.5 and 2.0 At inference when...
А где можно прочитать про минимальные/рекомендованные системные требования для генерации на 712x712 и 1024x1024 ?
    ```py import torch import numpy as np from PIL import Image, ImageOps from diffusers import AutoPipelineForInpainting import rembg modelName = 'kandinsky-community/kandinsky-2-2-decoder-inpaint' pipe = AutoPipelineForInpainting.from_pretrained(modelName, torch_dtype...
Noticed a yellow tint on most 2.2 images. image looks like sketches and illustration. Image quality deteriorates, poorer than 2.1.
I have a masked image and need to generate content for the mask region according to a depth map. Is there a model that can do inpainting and controlnet-depth condition...
Code to reproduce: ``` from kandinsky2 import get_kandinsky2 model = get_kandinsky2('cpu', task_type='text2img', cache_dir='/tmp/kandinsky2', 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=256, w=256, sampler='p_sampler', prior_cf_scale=4, prior_steps="5") ```...
Hi! Just want to highlight that there is no owner in the license file at the moment https://github.com/ai-forever/Kandinsky-2/blob/main/license#L189 
I am very interested in training a new controlnet model. After studying the kandinsky-2-2-controlnet-depth model uploaded in HuggingFace, I found that its architecture seems to be different from the controlnet...
Hi, Great work with the Kandinsky model, the last improvements look really impressive 🎨 For prior training/tuning I saw that the default batch size is 1, is that actually the...
I know this question doesn't really fit here, but anyway: Since https://editor.fusionbrain.ai/ has got a newer Version of Kandisky-2, the pictures have a totally different style and look. Is there...