Grid Splitter
Production friction turned into an open-source tool in a day
project
Approach
I used Gemini AI Studio to build a web app that handles it automatically. Upload a grid, the tool detects the layout, and extracts every tile individually. Added controls for dark/light line detection, sensitivity adjustment for different image styles, and edge trimming for messy borders. Download tiles one by one or grab them all as a zip.
Built and shipped the whole thing in a day. Deployed on Vercel, open-sourced on GitHub.
Value
What was eating 10–15 minutes per batch now takes seconds. The tool got picked up on the ComfyUI and Stable Diffusion subreddits immediately — turns out everyone working with grid-based outputs had the same problem and nobody had built the fix yet.
Challenge
Working on the PPG Above and Beyond campaign, I was generating hundreds of images through Nano Banan Prof — which outputs 3x3 composite grids. Every single one had to be manually cropped into individual tiles before they could be used. It was slow, repetitive, and killing iteration speed on a project that needed to move fast.
