I built an AI thing for Amazon sellers, and honestly it works way better than I expected

Decker Shaw — Tue, 09 Jun 2026 18:06:56 +0000

So here's the thing. If you sell on Amazon, you already know this pain: good product photos cost a stupid amount of money.
Like, a single lifestyle shot from a real photographer? That's $50 to $200 easy. A full listing photoset? You're looking at $500-$1000. And if you're launching multiple products a month like I was trying to do, that math just doesn't work.
So I spent the last few weeks building something. Here's what happened and what I learned along the way.
The problem nobody talks about
Most Amazon sellers end up in one of three situations:

Hire a pro photographer. Looks amazing, but $200 a shot kills your margins if you're selling anything under $30.
Phone camera DIY. Free, but honestly? It looks amateur and your conversion rate will show it.
Stock photos. Fast but Amazon's pretty strict about it and you might get your listing flagged. There had to be a better way. So I figured, why not just throw AI at it? What I built It's basically a pipeline. Three models working together, and the user doesn't even need to know which one is doing what: Seedream 5.0 Lite — main image generator. Costs about ¥0.22 per image. That's like 3 cents. Qwen-Image-Plus — backup if Seedream is being slow or rate-limited. DeepSeek V4 — handles the prompt engineering. You just say "a red coffee mug on a wooden table in the morning" and it figures out the rest. The workflow is stupid simple: describe your product in plain English -> AI understands what you want -> generates the image -> done. No prompt engineering degree required. Let's talk money Say you need 50 product images this month. Here's what that looks like: Pro photographer: $500 - $2,000. And you'll wait a week. Fiverr freelancer: $150 - $500. Still takes a couple days. My AI thing: $1.50. Yes, one dollar fifty. And it takes like 2 minutes. I had to double-check the math when I first ran it. I thought something was broken. Nope, that's just how cheap the API calls are these days. Things that went wrong Okay I'm not gonna pretend everything was smooth. A few things I wish I'd known upfront: Payment setup is a nightmare. Lemon Squeezy + PayPal was the path of least resistance. Stripe would be cleaner but getting a Stripe account without a US passport is basically impossible. New service = trust issues. People don't know you, they don't trust your AI. I'm dealing with this by offering a free tier — 3 images no credit card needed. Seems to help. Don't over-engineer. I literally almost spent two weeks building a full analytics dashboard before I had a single user. That would've been so dumb. The only question that matters is: can someone get a good image in under 60 seconds? Everything else is bonus. What's next Right now I'm working on: • Better quality on weird product shapes (smooth edges on irregular objects is apparently hard) • Bulk generation — upload a CSV with 100 products and get 100 images back • AI listing copy to go along with the images The service is live at deepcutapi.com if you wanna try it. Free tier, no card needed. Bottom line AI image generation for Amazon sellers isn't some future thing anymore. It works right now, and it's genuinely 90% cheaper than traditional photography. If you're still paying $500 a month for product photos, you should at least try the free tier and see if it works for your products. Drop a comment if you've tried AI tools for your Amazon listings. I'm curious what's working for other people. — Decker Shaw

DEV Community: Decker Shaw

I built an AI thing for Amazon sellers, and honestly it works way better than I expected