Michal Stary

prof_pic_old.jpg

I build and ship generative models for video, images, and 3D — from training to solving the last-mile research problems that make them ready for production at scale.

I’m an Applied Scientist at Zillow, where I lead video generation and image editing pipelines serving millions of users. Previously, I was a visiting researcher at MIT’s Scene Representation Group (supervised by Prof. Vincent Sitzmann), working on world-consistent video and multi-view diffusion models and sparse 3D reconstruction.

I hold an MSc in Computer Science from TUM (GPA 1.0, top 1%) and I am proudly part of the Technology Management Honors Degree from CDTM.

news

Feb 01, 2026 Our paper on realtime diffusion on 3DGS: GaussFusion has been accepted to CVPR 2026! arXiv
Jan 01, 2026 Our paper on generative video: Generative View Stitching has been accepted to ICLR 2026! arXiv
Oct 10, 2025 Presented our paper on understanding multi-view transformers at the End-to-end 3D Learning Workshop at ICCV 2025!
Aug 15, 2025 Our paper on understanding DUSt3R has been accepted to the End-to-end 3D Learning Workshop at ICCV 2025! arXiv
Jul 01, 2025 Graduated from TUM with an MSc in Computer Science (GPA 1.0, top 1%)!
Jun 01, 2025 Started as an Applied Scientist at Zillow in Munich to push the boundaries of cool real estate tech with generative models!
Nov 17, 2024 Won 2nd place overall at MIT Energy & Climate Hackathon 2024!
Sep 01, 2024 Started my visiting student researcher stay at MIT Scene Representation Group!
Aug 01, 2024 Concluded research internship at Virtual Staging AI (acquired by Zillow)!
Jul 15, 2024 Awarded Rohde-Schwarz Scholarship!
Apr 01, 2024 Concluded research internship (IDP) at SE3 Labs, check results here!

recent publications

  1. gaussfusion.png
    GaussFusion: Improving 3D Reconstruction in the Wild with a Geometry-Informed Video Generator
    Liyuan Zhu, Manjunath Narayana, Michal Stary, Will Hutchcroft, Gordon Wetzstein, and Iro Armeni
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
  2. Generative View Stitching
    Chonghyuk Song, Michal Stary, Boyuan Chen, George Kopanas, and Vincent Sitzmann
    In International Conference on Learning Representations (ICLR), 2026