Try it live Description
Vaho is a window for someone having a hard time far away: homesick, in a hospital bed, or just on the other side of the world. You wipe a fogged window with your finger and a place appears behind it. You draw on the fog, write a few words, and Vaho folds the place, your drawings and your message into a short link. Whoever opens it finds their name on the glass; a breath fogs it, your message writes itself, and when they wipe the glass, everything you drew is waiting outside for them.
There are eight windows: a town beneath a volcano in Ecuador, a rail town in Japan, a farm in Iowa, a courtyard in Provence, an island above the clouds, a greenhouse on the Moon, a garden under the sea, and a forest that lights up at dusk. In some of them you can climb through the frame and walk outside.
Drawing is the interface. Symbols change the place: a heart sends up paper lanterns, a star brings the night, a bird releases paper cranes. Animals can be drawn freely: a small neural network in the browser, trained on Google's Quick, Draw! dataset, recognizes ten animals, and a folded-paper version steps out and walks on the world's real ground.
Every view is a World Labs world generated from a single image (the source images were made with ChatGPT image generation), streamed as Gaussian splats with Spark behind a fog shader, with real parallax from the mouse, or from your head if you turn on the camera. Each world's collider mesh is the ground that creatures and visitors walk on, and the paper creatures stop at walls.
It runs in any browser, on phone or desktop, with nothing to install. Everything was made from scratch during the hackathon: the idea, the code, the eight World Labs worlds, the Tripo creatures and sill objects, the sketch recognizer, the sound and the walkthrough video. The music and sound effects were generated with code.
How Tripo was used
Tripo is the engine of the bestiary. In the author's words: without Tripo, a doodle would stay a doodle.
- Doodle to creature: a plain sketch goes through image-to-image to become a folded-paper version that keeps the sketch's pose, image-to-3D (H v3.1) builds the model, and auto-rigging plus retargeting give it a walk or swim cycle.
- Output: ten animals and the window-sill objects, all generated offline by the author, so the page never spends Tripo credits and visitors don't need to sign up.
- In the piece: when a visitor draws one of the animals on the fog, its paper version steps out and walks on the World Labs ground.
