T-Shirt Design AI

T-Shirt Design AI turns one sentence into a production-ready T-shirt file. The problem: AI image tools hand you a picture, but a picture isn't a print file. Anyone selling print-on-demand still had to drag the output into Photoshop to clean up edges, fix resolution and knock out the background before a printer would accept it. How it works: describe the design in plain English — subject, style, mood, slogan — or upload a reference image. In 60-90 seconds you get artwork previewed live on the shirt, switchable across 6 garment types and 12 colors in the browser, in 9 art styles (vintage retro, bold typography, cartoon mascot, line art, grunge streetwear, kawaii chibi, emblem badge, watercolor, Y2K). Export is a transparent 300 DPI PNG at 4500x5400, sized for DTG and screen print with safe margins, background removed by a semantic segmentation model. No watermark. Studio mode batches up to 20 designs in parallel from a list of slogans, and refunds failed generations automatically. Stack: Next.js 16 + React 19 + TypeScript, Tailwind CSS 4, Drizzle ORM on Neon serverless Postgres, NextAuth for Google sign-in, sharp for image processing, S3 for storage, Stripe for credit billing, all on Vercel. Generation runs long enough that Vercel's streaming and function timeouts saved me from building my own queue. First design and its print file are free after Google sign-in, no card needed.