
Kandinsky
Sber's official image-generation bot: describe a picture in Russian or English and Kandinsky draws it, mixes styles, and outpaints beyond the original frame — all for free.
Key features
- Free text-to-image generation
- Strong understanding of Russian-language prompts
- Style mixing between two images
- Outpainting beyond the original frame
Kandinsky
Kandinsky is Sber's own text-to-image model, and this bot is its official home in Telegram — the profile states plainly it is "Kandinsky by Sber AI." That corporate backing matters in a category full of anonymous wrapper bots: Kandinsky is a real, independently-trained model with its own research team behind it, not a reseller of someone else's API key. What sets it apart from most Midjourney-style bots is that it was built and tuned with Russian prompts as a first-class citizen, not an afterthought translated from English. Type a description in plain Russian and the model tends to understand nuance, local references and idioms noticeably better than image models built primarily around English datasets. It also supports style mixing between two images and outpainting — extending an existing picture beyond its original borders rather than only generating from scratch. With over 32 thousand monthly users, it has a solid, stable base. It won't match the absolute photorealism ceiling of the newest Western frontier models on every prompt, but for free, VPN-free, Russian-friendly image generation backed by a company that isn't going anywhere, Kandinsky is one of the safest choices in the category.
How to start
- Open @kandinsky21_bot and press Start
- Type a description of the image you want
- Wait for the picture and request variations if needed
Commands
/startOpen the bot/helpShow usage tipsPricing and limits
Free
Pros / Cons
- Official model from a large, stable company
- Excellent Russian-language prompt understanding
- Completely free
- Photorealism can lag behind the newest frontier models on some prompts
- Exact daily generation limits aren't published
Score breakdown
| Criterion | Weight | Points | |
|---|---|---|---|
| Functionality | 25 | 8 / 10 | |
| Reliability & speed | 15 | 9 / 10 | |
| Popularity & trust | 15 | 7 / 10 | |
| Price & free tier | 10 | 10 / 10 | |
| Usability | 10 | 8 / 10 | |
| Safety & privacy | 10 | 8 / 10 | |
| Freshness | 5 | 8 / 10 | |
| Localization | 5 | 10 / 10 | |
| No intrusive ads | 5 | 9 / 10 | |
| Score | 100 | 83.0 |
Alternatives
All alternatives →
БананоГенAssembles images from text in seconds using the Nano Banana model, and can also edit, upscale, remove…Нано Банана 2Upload a photo, describe the edit you want, and this bot applies it using Google's Nano Banana image model —…
Stable WaifuA dedicated anime and manga-style image generator with over three years online and nearly 70,000 monthly…
GigaChatSber's official free neural network bot: writes and explains text, summarizes long articles, draws pictures,…
Syntx AIA large multi-model creative bot with over 320,000 monthly users, offering 16+ image and video models…FAQ
Is Kandinsky really from Sber?
Yes, the bot's own profile identifies it as "Kandinsky by Sber AI," Sber's official text-to-image model.
Does it work well with Russian prompts?
Yes, it was trained with Russian as a primary language, not just translated support.
Technical details
- Inline mode
- Mini App
- Works in groups
- Works in channels
- Accepts voice
- Accepts files
- Requires phone number
- Requires registration
- Shows ads
- Works in Russia without VPN