Work AI product generator

From a picture to a priced, ready-to-make product

An AI product generator and shop: a customer uploads an image, refines it in a chat with an AI model into a manufacturing-ready SVG and saves it as a priced product, paying for generations with prepaid credits.

Sector
E-commerce and AI · Client project
Our part
First version, end to end, AI generation flow, Payment routing, Codebase foundations and tests
Built with
TypeScript, Next.js, React, Supabase, Vercel AI SDK, OpenAI API, Polar, Vercel
Service
MVP development →
Concept illustration · Fictional identity and sample interface, not the original product. AI-generated artwork.

Follow an image into a priced product.

The first version connected AI generation to a product a customer could save and price.

  1. 01Start with an image

    A verified account uploads an image and starts refining the idea in a chat with the model.

  2. 02Refine the design

    The model streams structured SVG output. Generated SVG is sanitized before it is stored.

  3. 03Save a priced product

    The customer approves the design. The first version priced the result from a fixed markup and its model cost.

  4. 04Support the purchase

    Prepaid credits support generation. Checkout and credit purchases are routed according to the buyer's billing country.

A simplified walkthrough of the project scope. No client data or original interface.

The engineering decision

Treat generated output as application data that needs validation and access control. Sanitize the SVG and keep each customer's records scoped to their account.

Our part, clearly defined

Our contribution was the first version end to end and later work on the codebase and payments. The owner's team subsequently developed the broader shop.

The starting point

The founder wanted customers to turn any picture into a product that can actually be made, and to buy it without waiting for a manual quote. That needs AI output that is safe to send to production, a price for every result, and accounts that keep each customer's work private.

What we were responsible for

We built the first version end to end and later reworked the codebase and payments. The product owner's own team develops it further.

What we built

  • The first version in one build

    Sign-up with email verification, image upload, a chat with the model that streams a structured SVG back, approval into a priced product and a personal dashboard.

  • Safe AI output

    Generated SVG is sanitised before it is stored, and every record is scoped to its owner with row-level security in the database.

  • A price for every result

    The first version priced each product from a fixed markup plus the model cost of making it.

  • Payments by country

    Checkout and credit purchases are routed to a local bank or an international payment provider by the buyer's billing country, with a switch to take a provider offline.

  • Foundations for a growing team

    A database client typed from the schema, one validation approach, server actions split by resource, translations in dictionaries, unit tests and one linter and formatter.

The result

The product grew from this base into a full shop — catalogue, cart, orders, credits and an admin area — developed further by the owner's team.

Working on something similar?

Book a 30-minute call or send a brief. If the scope is clear, you get a written fixed quote within one or two business days.