I built an AI portrait factory and listed it on Etsy
Five AI portrait listings went live on Etsy in a couple of hours, and the whole thing costs 15 cents per image to run.
I am going viral on TikTok. Fourteen likes on my Day 01 post. Basically a celebrity at this point. I should probably hire a manager. Anyway, I did not spend much time on that today because I had a better idea, one that might actually make money instead of feeding an algorithm that does not know I exist.
The idea itself is simple. People on Etsy are selling custom AI portraits of kids as popstars, astronauts, esports champions. They have hundreds of reviews and thousands of sales. It is saturated, sure. But saturated means there is demand. And with the right automation, the barrier to entry drops to almost nothing. I decided to find out if a couple of hours of work could get me a slice of that.
I built a pipeline that does almost everything for me. An MCP server plugs into my image generation API, five different agents each handle one portrait style (popstar, kpop, astronaut, ice hockey player, esports champion), and another agent writes Etsy descriptions and creates draft listings. The whole thing costs about 15 cents per image to generate and I can sell each one for 5 to 7 euros. The math is not complicated.
Is it AI slop? Probably. Do I care? Not really. People are making money doing this, and nobody is forcing anyone to buy it. Some parents want a picture of their kid as a K-pop star, and I am happy to be the middleman between them and a GPU. The quality looks good too, not the watermarked low-res stuff you get from free tools. That is the actual value proposition: you could prompt Gemini yourself, but it will look worse and take more effort than you expect. I did the work so you do not have to. That is worth five euros to someone, I think.
The bigger picture matters more than this one project. If the workflow works, I can reuse it for other automated Etsy shops on future days. The infrastructure is the real product here. A hundred listings across ten niches, all feeding into the same pipeline. That is not delusional, that is just repeating what already works. Some stores on Etsy do six figures with digital downloads. I am not saying I will get there with custom portraits. I am saying the pattern is proven.
Tomorrow I build something that has nothing to do with Etsy or AI portraits. The point of this project is variety, not depth. One new thing, done in a few hours, shipped before evening. If a portrait order comes in between now and then, the pipeline handles it on its own. That is the whole idea.