In the custom objects market (jewelry, furniture,etc.), there exists a significant disconnect between clients with specific visions and artisans seeking meaningful, personalized projects. Buyers often struggle to translate their vision into visual references due to limited artistic or technical knowledge, while artisans face challenges finding clients who value custom work and accurately communicating preferences. We want to empower anyone to visualize their creative ideas through playful AI interactions, then connects them with skilled artisans who can bring those visions to life.
Big thanks to our mentors Trudy Painter, Kapil Shukla, Sophia Sterling & Renee Zhang from the Google Team on helping us bring this project to life.
Traditional AI image generation interfaces like Midjourney often overwhelm users with open ended, complex and text-only inputs. But with our product, users start their journey by a guided prompt to articulate intent for object, occasion, and who the object is for. Users respond better to natural language than technical specifications, while guided contextual information (occasion, recipient) helps AI understand style preferences.
Nine designs are displayed simultaneously, and users eliminate unwanted options by tapping “whack-a-mole” style teaching the AI through actions over words. Then new options generate based on rejected designs.
Negative selection (rejection) is more intuitive than positive selection. Game-like interaction reduces decision fatigue & cognitive load. Through this method, there is higher user active participation over passive browsing
Active Preference Learning
The system observes patterns in user decisions (e.g., avoiding silver jewelry), proactively confirms detected preferences, and builds a dynamic "taste profile" that evolves with user interactions
Users want visibility into AI's understanding thus manual override options are essential for trust. an organized preference display helps users refine their vision, and separating likes/dislikes creates clearer decision boundaries. A dynamic tab displays learned preferences where users can manually add/remove preferences.
Artists can find projects they want to take on and bid on them.
After users finalize their AI-generated design, they can post it to the marketplace where artisans can discover it through intelligent filtering based on their craft specialties. Instead of relying on chance encounters, the platform actively suggests relevant projects to artisans based on their portfolio and past work, ensuring that skilled craftspeople see opportunities that match their expertise. Once connected, artisans can bid on projects they're excited to create, establishing a direct line of communication with clients.
We interviewed 5 artisans, and they strongly prefer seeing fully realized design concepts rather than trying to interpret vague client descriptions, making the AI-generated visualizations invaluable for clear communication.
Bridge Building
Technology serves as an enhancement to the creative process, not a replacement for human craftsmanship. The platform creates a clear handoff point between AI-assisted visualization and human creation, ensuring that both sides of the equation retain their value. This balance maintains the integrity of artisan work while using technology to solve the visualization challenge.
