What I Asked For
The starting concept was a generative art app where typing text produces animated visuals driven by two things simultaneously: the semantic weight of the words themselves, and the physical shapes of the letters. Soft, slow words like “dream” or “still” would generate ambient, trailing motion. Fast, hard words like “fire” or “shatter” would produce something fragmented and strobing.
From there, the priority became the share flow. Not just a visual you look at alone, but something you could send: type a message, hit Share, and the recipient gets a full-screen, chrome-free version of your living letter at a unique URL. The URL format is a poetic adjective-noun slug generated from the word pool in the app, something like “wandering-fire” or “silent-bloom,” so even the link itself carries a little bit of the feeling.
What It Is
Living Text is a full production web app, not a self-contained file. It runs on Next.js, deploys on Vercel, and uses Vercel KV (Redis) to store shares with a 30-day expiry. When you hit Share, an API route saves your text and returns a slug. That slug resolves to a viewer page that shows nothing but the visual, with a subtle “create your own” link in the corner.
The visual engine itself is a dual-canvas system with eight motion forms: pulse, burst, spiral, bloom, wave, fall, shatter, and bounce. The semantic analysis happens client-side, mapping words and letters to color ranges, motion types, and intensity levels. No model calls, no backend processing. The meaning lives in the word map.
It’s at livingtext.vercel.app. Type something slow and then type something fast and see what happens.