ant-research/4DAnyone

[SIGGRAPH Asia 2026] 4DAnyone: Create Anyone in 4D from a Casual Monocular Video

What it solves

4DAnyone transforms a single, casual monocular video of a person into a set of multi-view videos. This solves the problem of needing expensive multi-camera rigs to capture 4D (3D + time) human reconstruction, allowing users to generate a dense set of target views from a simple portrait video.

How it works

The system takes a portrait video (720p or higher) of a person and generates multiple synthetic views of that person across different camera configurations. Users can specify the number of views per layer, pitch angles for different height levels, and the horizontal yaw span to create a custom camera rig. The output includes generated target view videos and camera metadata, which can then be used as input for downstream 4D Gaussian Splatting (4DGS) reconstruction.

Who it’s for

This tool is designed for researchers and developers working in 4D human reconstruction, computer vision, and digital human creation, as well as those interested in integrating synthetic multi-view data with 4DGS pipelines.

Highlights

  • Flexible Camera Layouts: Supports various configurations, from a simple 6-view orbit to a dense 48-view layout across three pitch layers.
  • 4DGS Integration: Provides scripts to export data for Nerfstudio and train foreground-only 3D Gaussian Splatting models.
  • Customizable Viewpoints: Allows precise control over the number of views, pitch, and yaw coverage.
  • Automated Workflow: Automatically downloads required models and examples upon first use.

Related

  • Project
    zju3dv/Diffuman4D

    Diffuman4D is a spatio-temporal diffusion model for high-fidelity 4D consistent human view synthesis, enabling free-viewpoint rendering from sparse-view videos.

  • Project
    IamCreateAI/NeoVerse

    NeoVerse is a 4D world model that reconstructs 3D scenes from monocular videos to enable high-quality video generation along novel camera trajectories.

  • Project
    Eyeline-Labs/Vista4D

    Vista4D is a video reshooting framework that uses 4D point clouds to synthesize dynamic scenes from source videos using novel camera trajectories and viewpoints.

  • Project
    yehonathanlitman/Lift4D

    Lift4D reconstructs temporally consistent 4D assets from single-view in-the-wild videos by fusing per-frame 3D estimations into a deformable model.

  • Project
    facebookresearch/mvdust3r

    A single-stage 3D scene reconstruction tool that creates 3D point clouds and camera poses from sparse RGB views in about 2 seconds without requiring pre-known camera poses.