AI Model2026
Turns natural-language descriptions into runnable Three.js scene code, exploring how a small language model can generate structured interactive 3D environments as well as general-purpose code.
AI Model2026
Turns a single object image into a viewable 3D Gaussian splat, generating up to about 1.05 million Gaussians as a standard PLY file. It accepts photographs, renders, and generated images.
AI Video2026
Generates short videos with synchronized stereo audio from prompts, endpoint frames, or multimodal references. Six-step sampling targets faster creative iteration; the hosted API adds higher-resolution output and separate audio stems.