Back to the red bull cans

Co-built with Jay Prakash Pathak

Vacroni

From a text prompt to a canvas art product, ready to sell

ShelvedJan 2026

Video in the works

Vacroni

One prompt becoming five product photos and a Shopify listing.

About the project

Vacroni turns an idea for wall art into a product ready to sell. Type an idea, and the backend makes the art, photographs it in five settings, writes the listing and uploads it to Shopify.

Vacroni turns an idea for wall art into a product that's ready to sell. Type a prompt: the backend makes the artwork, photographs it in five settings, writes the listing, and uploads it to Shopify.

Jay Prakash Pathak and I built it in January 2026. Only the backend exists; we shelved it half built.

How it works

  1. Prompt

    Describe the artwork.

  2. Master image

    One master front view is generated.

  3. Four more shots

    The master is edited into a close-up, the frame edge, the canvas texture, and the size on a wall, so the art is identical in every photo.

  4. Listing

    AI writes the title, descriptions, price and tags.

  5. Shopify

    The product goes up with its images, price and discount.

Under the hood

API
Express and TypeScript on Postgres.
Images
FAL nano-banana-pro and sharp.
Text
Google Gemini.
Store
The Shopify Admin GraphQL API, plus CSV exports.

What I did

  • An API to generate, list and upload products
  • One master image, then 4 edits of it: close-up, frame edge, canvas texture, size on a wall
  • AI-written titles, descriptions, prices and tags
  • Shopify upload with images, prices and discounts, and CSV exports
  • Market research on competitors and Indian art categories

What was new

  • Editing one master render, so the art is identical in every photo

Problems I hit, and how I fixed them

  1. Separate renders gave a different artwork in each photo.

    Fix Make one master, then edit it for each angle.

  2. Image credits ran out mid-run.

    Fix Catch it and show a clear error.

Screenshots

Screenshots in the works

Real screens from Vacroni go here.

To be clear

There's no git history, so how the work split between Jay and me can't be shown.