One file in, any product out - automating a fashion factory

A file-ingest engine that turned one master graphic into print-ready garment files for every product and size - errors near zero, 15 printers fed nonstop.

Art Lethe manufactured for print-on-demand fashion brands - Mr. Gugu & Miss Go, Live Heroes, Carpatree, Aloha From Deer - where hundreds of new products launched daily and any design had to become print-ready files for every garment type and every size, XXS to 4XL. Done by hand, that meant thousands of files a day and a constant stream of production errors. As Head of the Graphics Department, leading a team of three designers, I set out to make the manual work disappear.

Over two years I designed and refined a fully automated file-ingest engine: drop a single 100x100 cm master graphic, and the system splits it into garment objects - front, back, sleeves, cuffs, neck, per size - named, validated and auto-laid-out for 15 sublimation printers and 2 calenders. Underneath sat work nobody sees: 300+ cutting templates vectorized from hand-made patterns, 10,000+ base graphics cleaned and unified, a file-naming dictionary (brand / date / SKU / mold / gender / size / object) and a product index taxonomy - all wired into the in-house CMS and generation server through Illustrator and Photoshop scripting: actions, droplets, batch pipelines.

The result was a complete reformation of how the factory handled graphics. Production errors fell to near zero, the entire product database was rebuilt on the new taxonomy within two years, and the pipeline generated flawless files at a scale no team could match by hand. The brands it powered still sell today. It remains the project I measure others against - and the direct ancestor of the AI-driven pipelines I build now: same discipline, new tools.