Atlas generates full 3D environments from a spoken description and lets you reshape them in real time through conversation.
Voice commands can generate worlds, add objects, adjust lighting, and restyle surfaces. The resulting scene is rendered as persistent, explorable 3D geometry in the browser.
Built with Yehor Ivanenko at the Conversational AI Hackathon hosted by Moss (YC F25) at Y Combinator.
Features
Voice-to-3D world generation
Conversational scene editing and object placement
First-person browser exploration
WebXR augmented reality support
Built with JavaScript, Python, Three.js, WebXR, LiveKit, Qwen, Hunyuan, 3D Gaussian Splatting.
