AAI News Hub
ResearchThu, August 20, 2026·1d ago

4DAnyone reconstructs 4D humans from monocular video

The framework uses multiview-consistent video generation and 4D Gaussian Splatting.

Why it matters

The work targets a key limitation in human 4D reconstruction: maintaining global appearance and structure across many generated views. If validated, it could improve pipelines that turn casual video into editable or renderable 4D human assets.

The key points

  • 1.Reconstructs 4D humans from uncalibrated monocular video.
  • 2.Uses Reference Context Packing and Target Context Routing.
  • 3.Targets consistency failures across many generated views.

Researchers presented 4DAnyone, a framework for reconstructing 4D humans from uncalibrated monocular video. The method generates reconstruction-grade, multiview-consistent videos and lifts them into 4D Gaussian Splatting, addressing consistency failures in camera-controlled video diffusion when many target views are required.

Try this today

Read the paper before using camera-controlled video diffusion for multiview human reconstruction.

Sources & original reporting

This brief summarizes and links to reporting from the publishers below.

Enjoyed this brief? Get the next one in your inbox.

More in Research