# ai-forever/Kandinsky-2

Kandinsky 2 — multilingual text2image latent diffusion model

Repository: https://github.com/ai-forever/Kandinsky-2
Canonical: https://ross.abutalabs.com/products/kandinsky-2
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Topics: image-generation, text-to-image, diffusion, image2image, inpainting, ipython-notebook, kandinsky, outpainting, text2image
Last push: 2024-05-01T17:03:31+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1419, "days_push": 854, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2814, forks 319 (observed 2026-08-28T04:07:23.815071+00:00)

## What it is
Kandinsky 2 is a multilingual text-to-image latent diffusion model library from Sber AI, supporting text2img, img2img, inpainting, outpainting, and ControlNet-depth generation. It ships PyTorch checkpoints on Hugging Face with Jupyter notebook examples and LoRA fine-tuning support.

## Use cases
- generate images from text prompts
- multilingual text-to-image generation
- inpaint missing parts of an image
- outpaint to extend an image beyond its borders
- image-to-image transformation guided by text
- control image generation with depth maps via ControlNet
- fine-tune a diffusion model with LoRA

## When to choose
- you need a diffusion model that understands non-English prompts
- you want text2img, inpainting, and outpainting in one model family
- you want open Apache-2.0 checkpoints runnable on a single CUDA GPU

## When to avoid
- you need the newest image quality from actively developed models like SDXL or Flux
- you have no GPU available
- you need a managed API rather than self-hosted inference

## Facets
- artifact type: library
- maturity: maintenance
- function: image-processing, machine-learning, deep-learning
- domain: image-processing, artificial-intelligence, deep-learning
- platform: python
- tags: text-to-image, latent-diffusion, inpainting, outpainting, controlnet, multilingual, stable-diffusion, gpu

## Member repositories
- ai-forever/Kandinsky-2 (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:23.815071+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T08:13:28.446229+00:00, confidence not recorded.
  - readme: https://github.com/ai-forever/Kandinsky-2 (fetched 2026-08-28T04:07:23.815071+00:00, sha ddc110e3353a)
- Data as of 2026-08-30T08:39:29.467469+00:00.