Avant Sanjhi

What it does

Avant Sanjhi helps revive Mathura’s dying Sanjhi craft by combining generative AI with traditional artistry to reduce design times and empower artisans to create personalized modern interpretations that make the tradition accessible, relevant and market-ready.

Your inspiration

Sanjhi, a dying 400-year-old Indian paper-cutting craft, has dwindled to 2 families, losing its voice to architect-commissioned stencils for financial sustenance. Artisans were reduced from image makers to glorified stencil cutters. Seeing craftsmen lacked tools for creative growth to adapt and struggle to modernise imagery, I saw potential for an AI-powered creative copilot to spark new visual language, bridging ancient techniques with modern aesthetics. The solution was two-part: restore artisans' creative agency, while opening Sanjhi to a global gift-and-craft market so that they could earn while learning patron preferences directly.

How it works

Users upload any meaningful image -objects, flowers, or places to the web platform. Four specialised AI agents built on ChatGPT-4o and DALL·E 3 APIs transform the image into Sanjhi-style designs. The analysis agent identifies composition and key elements. A moderation agent then filters content against Hindu principles, rejecting images with violence, animal cruelty, or vices like smoking. A stylization agent applies carefully engineered prompts that understand Sanjhi's core traits: symmetry, flatness, central focal point, and the 'Nas' - flowing negative space. Finally, a generation agent encodes these craft rules into refined text-to-image instructions to produce the modern Sanjhi interpretation. A MongoDB backend tracks user data to improve prompt performance. This process reduces the sketch phase from 2–8 weeks to days, giving artisans faster turnaround, fresh design possibilities, and digital guides that still require skilled hands to bring them to life.

Design process

Interviews with Ashutosh Verma (6th-gen artisan) and Sakshi Garg (apprentice of Master Ram Soni) revealed Sanjhi’s stagnation in commercial commissions and imagery. The idea of AI integration was met with resistance and fear of replacement. I developed two-stage prompts (analysis → styling), capturing nine metadata points to decode Sanjhi’s visual language and test if AI could replicate the style. Development of an early ML stencil prototype quickly revealed Sanjhi wasn't about stencil-making, it's uniquely stylized personal storytelling. A four module agent was built using OpenAI’s Python library; ran 460+ API calls and 220+ image generations. Added a moderation agent rooted in Hindu dharmic values to honor the craft’s temple origins. Launched a public site; collected user and artisan feedback via Google Forms and MongoDB. Tested 70+ image types. Additionally, failed Zund cutter tests proved that the artisans’ paper-cutting skill (<80gsm paper) and imagery expertise are irreplaceable. In co-creation trials, 3 AI-generated designs were interpreted by artisans into handcrafted Sanjhi works, blending tech input with their own stylistic mastery. Feedback loops informed prompt refinements, and real commissions alongside a London exhibition validated the design quality and concept.

How it is different

Avant Sanjhi is the first tool to use AI for craft conservation in India and the first ever digital tool built specifically for Sanjhi. A full web-to-tangible pipeline including real-time personalized design previews sets a new standard for custom craft commissioning. By using AI's limitations in understanding ‘Nas’(the vein-like flow), the tool invites artisans to reinterpret, redraw, and co-create, ensuring their individual creative agency dominates the final work. It reduces design time from weeks to days, while retaining complexity that keeps craftsmanship central. A dharmic moderation layer aligns creative AI with Hindu ethical codes to preserve the craft’s sacred roots in Hindu temples, making it the first of its kind. This platform empowers artisans to expand their creative vocabulary while their individual interpretation remains the defining force in every handcrafted piece, bringing heritage into modern markets through ethical, collaborative design.

Future plans

This platform could shift India's often rudimentary human-centered craft revival approach into one powered by accessible technology for artisan survival. I want to relaunch the web tool as a commercial website to test its viability with wider audiences. Develop a revenue-sharing model with artisans on these commissions. Future explorations would involve exploring AI as a potential aide in craft conservation of other 2D dying crafts like Gond art and Pattachitra. My real joy lies in making complex technologies like this API-powered tool, usable by common craftspeople who are intimidated by technology and unable to harness its potential.

Awards

Shortlisted for the Design Intelligence Award 2024 for Cultural Innovation. During free public access, an artisan used the tool for a Bangalore Craft Museum commission to depict the city’s skyline in Sanjhi style, reporting a 4-week reduction in sketch time and help in visualization of non-traditional, urban imagery.