I’m building Toonbead, a photo-to-fuse-bead-pattern tool for makers who want results they can actually build.
The new part is an AI restyle step between the uploaded photo and the bead grid.
Most photo-to-pixel tools treat the image as the final result. That is not enough for fuse beads. A craft pattern needs a readable subject, broad color regions, a manageable palette, and enough structure to assemble on a pegboard.
Toonbead now splits the workflow into two jobs:
The API is deliberately narrow. Up a photo does not call the AI service. Changing the grid or colors does not call it again. One explicit Generate action makes one same-origin request, and the result is kept only in the active browser session so the local craft workflow can continue without another generation.
The account flow is also simple: verify an email to receive three free AI generations. After that, one-time generation packs are available; there is no subscription. The browser-local Basic path stays free, including local adjustments and exports.
The privacy boundary matters here. The uploaded photo is transiently proxied through the server to the image model during an explicit AI Generate, then the generated image is streamed back. Toonbead does not persist the original or generated image, image URLs, or prompts. That is different from claiming the photo never leaves the browser.
I’m still treating the AI output as a starting mother image, not a final artwork or a guaranteed likeness. The craft-specific work remains visible and editable: real bead colors, board counts, bead counts, cleanup, and a pattern a person can follow.
Try the current maker: [https://toonbead.com/maker](https://toonbead.com/maker)
If you build fuse bead art, pixel art, or photo-based craft projects, I’d love feedback on where this workflow can become more useful.