Memory Museum

Inspiration

Photo albums preserve what happened, but a keepsake can bring the memory back in a way a timeline cannot. I built Memory Museum to give meaningful objects a place of their own, with a room and a story written by the person who remembers them.

What I learned

AI is most useful here when its role is limited. Memory Lens can suggest what an object looks like, but it should not invent why it matters. Photos go to an external vision provider only after the user opts in, and the manual path still works without AI.

How I built it

Memory Museum is a Next.js and TypeScript app. React Three Fiber and Three.js render the rooms and procedural keepsake stand-ins. People start with an empty room, arrange furniture, add photos, and write a story. Room layouts, furniture rotations, and keepsake positions save to local JSON or PostgreSQL through Drizzle.

Challenges

Making floor clicks map reliably to room positions, then persist across reloads, took care. I also kept the visual models honest: they are editable procedural stand-ins, not claims of accurate 3D scans.