What they did
After seeing a video of a flower growing out of glowing characters, the poster wanted a tool that could do the same to any picture and asked Claude Code, running Fable 5.2, to build it. It began as a single prompt, followed by a few short requests such as adding templates and fixing long images that were being cropped. The result is one HTML file with no backend and no frameworks, built on WebGL2 and Web Audio, published on GitHub Pages with its code on GitHub.
How it went
The tool takes an image, GIF, video or typed word and renders it in nine styles, including terminal, halftone, newsprint and matrix, with custom character sets. It can turn the picture into a 3D object that finds the subject and keeps the background flat, with an in-browser depth model as an option. It also writes a short soundtrack from the image and exports PNG, GIF or video. The poster gives no numbers on time or cost.
Worth knowing
Everything runs locally in the browser, so nothing gets uploaded. The post calls it an experiment and does not say how many follow-up prompts it took beyond a handful or what broke along the way.