Image
omnigen-vault
Infinite text-free AI image generator with a searchable multilingual vault
★ 35 starsMIT
View guide →Kandinsky 2.2 — multilingual text-to-image latent diffusion with ControlNet, inpainting and image fusion
Kandinsky-2 is an open-source image project by ai-forever (ai-forever/Kandinsky-2 on GitHub). It has 2,810 GitHub stars and 319 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 1 May 2024.
Pick the task — text2image, img2img, inpainting or image fusion — and the model version (2.2 is the newest in the repo).
Clone the repo or pip install it directly from GitHub as the README shows; model checkpoints download from Hugging Face.
Set up a CUDA environment (examples call get_kandinsky2('cuda', ...)) and open the example notebooks in ./notebooks.
Call generate_text2img with a prompt — prompts can be multilingual — and inspect the output image.
Trained with multilingual text encoders (XLM-Roberta-based), so prompts work in many languages, not just English.
2.2 switches to the CLIP-ViT-G image encoder for more aesthetic results and better text understanding.
2.2 adds ControlNet (e.g. depth-conditioned decoding) for more controllable generation.
Dedicated inpainting checkpoint and generate_inpainting API with masks.
mix_images blends multiple images and text prompts with per-input weights.
A linked Colab notebook shows fine-tuning with LoRA, alongside an inference Colab.
Clone Kandinsky-2 and enter the directory.
git clone https://github.com/ai-forever/Kandinsky-2.git
cd Kandinsky-2Install directly from GitHub as the README shows.
pip install "git+https://github.com/ai-forever/Kandinsky-2.git"Use the README's text2image example with the 2.2 model on CUDA.
from kandinsky2 import get_kandinsky2
model = get_kandinsky2('cuda', task_type='text2img', model_version='2.2')
images = model.generate_text2img("red cat, 4k photo", decoder_steps=50, batch_size=1, h=1024, w=768)ai-forever (the team behind the Kandinsky models, originally from Sber AI research) maintains this repo; the authors are listed in the README.
The code is Apache-2.0 licensed; checkpoints are on Hugging Face. You need your own GPU hardware to run it.
No — inference runs locally in PyTorch once checkpoints are downloaded.
Both are open latent-diffusion models. Kandinsky is distinguished by its multilingual text encoders, diffusion image prior design, and built-in image-fusion workflow.
pip install it from GitHub, then run the text2image snippet or the linked Colab inference notebook.
Infinite text-free AI image generator with a searchable multilingual vault
MMagic — OpenMMLab's generative-AI toolbox: diffusion, text-to-image, restoration and a model zoo.
Create 🔥 videos with Stable Diffusion by exploring the latent space and morphing between text prompts
Desktop AI art studio — photoreal renders, ControlNet and upscaling
A lite C++ AI toolkit: 100+ models with MNN, ORT and TRT, including Det, Seg, Stable-Diffusion, Face-Fusion.
Drop-in React image editor — crop, filters, draw, text and AI assistant
The most-starred other Image projects in the LibreAI catalogue — the realistic shortlist if Kandinsky-2 is not the right fit.
Stable Diffusion web UI
The node-graph powerhouse for precise, professional AI image and video generation.
High performance self-hosted photo and video management solution
Focus on prompting and generating
🆙 Upscayl - #1 Free and Open Source AI Image Upscaler for Linux, MacOS and Windows
Free and open — grab the source, star the repo, and start creating.