Challenge
Image and video models became good enough for everyday use, but most people don’t want accounts with several providers, separate payments and the craft of writing prompts. The idea was one Telegram service where anyone can make an image or a short video in a chat.
Approach
I built and ran it myself in 2025–2026.
- Inside Telegram. A bot plus a Mini App front end at creato.space, so nobody has to leave the messenger.
- Models through APIs. Image generation with Nano Banana, video with Veo, and other models added over time.
- Prompt assistant. A GPT-based helper in the chat that turns a rough idea into a working prompt.
- Tools. Background removal next to generation.
- Money. Five free generations for every new user, then a balance topped up by payment or a subscription.
- Growth. Paid advertising, plus a Telegram channel with example images and ready-to-use prompts that led people to the bot.
Result
The service reached about 40,000 users a month. Paying users were a small share, and most of them used up what they bought.
The economics did not hold. Five free generations for every new user, paid at API prices, cost more than payments brought in, and there was no reserve to carry the loss through the early stage. I closed the service in 2026.
The lesson I took from it: work out the cost of the free tier per user before paying for traffic. People came and used it; the pricing didn’t hold. More about AI creative production.