
Sber's official free neural network bot: writes and explains text, summarizes long articles, draws pictures, and understands both text and voice messages — no VPN or subscription needed.
These Telegram bots turn a text prompt into an image using diffusion models, for illustrations, avatars, concept art, and quick edits without opening a separate app.

Sber's official free neural network bot: writes and explains text, summarizes long articles, draws pictures, and understands both text and voice messages — no VPN or subscription needed.

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.

Send a picture and this bot removes the background in seconds, returning a clean cutout with a transparent background — no editing skills required.
| # | Bot | Price | Languages | Inline | Groups | Score |
|---|---|---|---|---|---|---|
| 1 | GigaChat @gigachat_bot | Free | RU · EN | no | no | 85 |
| 2 | Kandinsky @kandinsky21_bot | Free | RU · EN | no | no | 83 |
| 3 | AI Background Remover @AI_Background_Remover_Bot | Free | EN | no | no | 68 |
We generated the same 5 prompts on every bot: a photorealistic portrait, a simple flat icon, a scene with text inside the image (most models still render this badly), a picture in a specific art style, and a request to upscale or edit an existing photo. We counted how many free generations the bot allowed before asking to pay, and timed how long each image took to render.
Look at which engine powers the bot, not just its sample gallery — a bot built on a newer diffusion model renders hands, text, and complex scenes far more reliably than an older one, even if both promise "photorealistic" results. Test with a prompt containing text you want inside the image, like a sign or a label; that single detail exposes a weak model faster than any portrait.
Almost every image bot runs on paid compute, so "free" usually means a starting credit balance — commonly 5–20 generations — that refills slowly or not at all. Check whether unused credits expire daily and whether upscaling or regenerating a variation costs a separate credit; some bots charge for each retry even when the result is worse than the first attempt.
Free tiers often cap output at a lower resolution, roughly 512–768 px on a side, and add a watermark, while paid tiers unlock full resolution and commercial-use rights. Before using a generated image publicly, check the bot's terms: not every service grants commercial rights even on a paid plan, and prompts referencing a living artist's name or a copyrighted character can get outputs blocked or flagged.
Yes — most bots give you a starting batch of free generations right inside Telegram with no separate signup, though it's usually capped at 5–20 images.
It's a limitation of the underlying model, not a specific bot — diffusion models generate text pixel by pixel rather than letter by letter, so signs and labels often come out garbled.
It depends on the bot's own terms and plan — not every service grants commercial-use rights even on a paid tier, so check before publishing.