Doppel is a cross-embodiment video world model from Johns Hopkins University, Carnegie Mellon University and Toyota Research Institute. Given an initial image, camera-aligned renders of a planned motion and its numerical trajectory, it predicts how the scene and objects will move.

Overview

Model type
World simulation
Architecture
Transformer
Listed versions
Doppel-8

Papers & reports (1)

Render the Robot, Generate the World: One Video World Model Across Embodiments

Paper available

arXiv: 2611.24925
Authors & affiliations (7)

Author order: Original order

  1. Yash Jangir

    Johns Hopkins University

  2. Karan Mirakhor

    Carnegie Mellon University

  3. Renos Zabounidis

    Carnegie Mellon University

  4. Muhammad Zubair Irshad

    Toyota Research Institute

  5. Sergey Zakharov

    Toyota Research Institute

  6. Yonatan Bisk

    Carnegie Mellon University

  7. Homanga Bharadhwaj

    Johns Hopkins University

First published
2026-11
Latest paper revision
v1
Venue
arXiv

Covered model versions

  • doppel · Doppel-8

    Shared render-action video world model and the eight-embodiment Doppel-8 experiment.

Paper ↗PDF ↗

Official resources

Submit a source or correction