Apple
Apple's single-photo → 3D model with view-dependent lighting and reflections
LiTo (Surface Light Field Tokenization) is an Apple ML Research model (ICLR 2026) that reconstructs a complete 3D object with view-dependent appearance — specular reflections, Fresnel effects — from a single RGB photo. It encodes the surface light field into compact latent vectors capturing both geometry and lighting. Unlike static 3D reconstructions, the output renders correctly as the viewer moves around the object. **Step-by-step:** 1) `git clone https://github.com/apple/ml-lito`. 2) Download `lito_new.ckpt` checkpoint from repo releases. 3) `pip install -r requirements.txt`. 4) Run the inference script with a single input image. 5) Export the 3D representation for use in game engines, AR apps, or VFX pipelines. Demos at apple.github.io/ml-lito.
Extracts a view-dependent 3D model with correct reflections and lighting from one photo.
Free (code + checkpoint on GitHub, Apple ML Research license)