Gemini Sketchbook Search

Year: 2024

Client: Google Creative Lab, New York, NY

Technology: Gemini API, multimodal embeddings, OCR, image captioning, browser-based JavaScript

Description

Experimental search and browsing interface for decades of Kyle McDonald’s sketchbooks, using Gemini to read, describe, and connect handwritten notes and images.

The prototype turns a private archive into a multimodal index: pages can be scanned, embedded, searched, and recombined through image and text recognition, making older sketches accessible as material for new research and demonstrations.