Ever see a cool AI photo transformation prompt on Twitter or Reddit and think “I’ll use that later,” then completely lose track of it? I built a tool to fix that for myself.
What it does
Reshaper is a local web app that’s basically a personal prompt library with a real interface on top. Paste any AI transformation prompt you find into an admin panel, tag it with a category, and it becomes a usable style in your own picker, upload a photo, pick a style, generate. Currently sitting at 32 prompts across 7 categories (Doodle & Sketch, 3D & Figure, Comic & Pop Art, Fine Art & Painting, and more), with a before/after slider to compare results and an option to view and tweak the exact prompt before generating.
Technical approach
No shared API key: everyone brings their own Gemini or OpenAI key, processed locally, never saved to disk. No billing infrastructure, no ongoing cost to me. Zero npm dependencies: native Node.js, HTML, CSS, and JavaScript, no frameworks, no build step. Style and category management: a local admin panel to add, edit, and organize prompts and categories without touching code, essentially a lightweight CMS for a prompt collection.
Key insight
Working with an AI coding assistant across many sessions surfaced the same lesson repeatedly: the value wasn’t generating code fast, it was having a collaborator that could hold the whole system in view, catching a duplicate style hiding under a different name, flagging a design inconsistency, or noticing a security tradeoff before it became a problem.
Personal project, no accounts, no billing, no service guarantees. Code’s open on GitHub: wayneo/reshaper
