Try it live Description
Carry the Cosmos is a gift for urban flâneurs that turns a pause on a walk into an AR encounter. Offer a steady palm or finger, and a bird approaches, perches and changes position; move your hand away to say goodbye. Pick a spot on the ground to plant a 3D plant, watch it unfold and walk around it. You can wait, daydream or stop at any moment: waiting and leaving are both part of the experience.
Under the hood, Apple Vision detects hand landmarks, and a rule engine uses perch position, duration and tracking state to sequence 2.5D bird animations. ARKit detects surfaces and tracks the device in space; plants pause their growth if tracking is lost.
Six plants are available at the start. Every 10,000 valid steps, accumulated across days, unlocks a plant of your choice, and any plant can be planted again. Each planting gets its own record and, when a location is available, a place on the map. Back at the original site, the same iPhone can relocalize from a saved spatial map and restore the plant where it was placed. Photo-generated World Labs scenes can be reopened from the map; merging them with individual AR plants is still planned.
The native AR version runs on ARKit-capable iPhones via TestFlight, and the web app previews the product interface. The author built the whole project alone during this Tripothon. Besides native AR gesture interaction, it covers Tripo character integration, urban planting, a journal, a circular hardware prototype, a photo-generated 3D world beneath a tree with companion interactions, map-based bird encounters, and separate Chinese and English entry points.
How Tripo was used
Tripo provides the 3D characters in Carry the Cosmos.
- Avatars and walking companions: the 3D models come from Tripo, and the engineering materials include the runtime GLBs and provenance records.
- Integration: the author brings the assets in through Three.js, calibrates scale and heading, refines animation and gait, and connects them to the map.
The AR bird uses a separate set of 2.5D animation assets, with gesture rules coordinating its responses.
