PHULKARI 2.0: GENERATIVE PATTERN SYSTEMS AND AI-DRIVEN EMBROIDERY FUTURES IN PUNJABI TEXTILE TRADITION

Phulkari, the vibrant hand-embroidery of Punjab, is fading as machine-made imitations inundate the market and the number of practicing artisans falls below 5,000. Phulkari 2.0 combines careful heritage documentation with human-guided artificial intelligence to transform that threat into an opportunity. We photographed and vectorized 1,218 historical and village-sourced motifs, then trained a conditional DCGAN that respects two cultural signals: the motif’s wallpaper-symmetry class and the natural-dye color range. With a 70 / 15 / 15 % split, the model achieved an FID of 18.2 and an Inception score of 3.74, far surpassing a VAE baseline (FID 45.1). A panel of 12 experts rated 73 % of generated designs as “authentic” (≥ 4 / 5). Field uptake is already evident: two women-led cooperatives have licensed 14 AI-generated motifs, reporting 1.8× higher profit margins and a 70 % reduction in manual drafting time. Oversight rests with a four-member Heritage Stewardship Board that reviews every new design. Training the model required 41 kWh of electricity—about 25 kg CO₂e—a footprint we monitor as part of our commitment to responsible AI. This work demonstrates how community-directed AI can revive endangered crafts while preserving cultural integrity and economic fairness.