AI People Wiki

Builders & open source

Dani Valevski

Co-first author of GameNGen, a diffusion model that simulates DOOM in real time, and first author of UniTune

Dani Valevski is a co-first author of "Diffusion Models Are Real-Time Game Engines" (2024), which presents GameNGen, a neural model that interactively simulates the game DOOM in real time by training a diffusion model to predict the next frame from past frames and actions. The GameNGen project page lists Valevski with Google Research. Valevski is also first author of "UniTune: Text-Driven Image Editing by Fine Tuning a Diffusion Model on a Single Image" (2022).123

At a glance

Current role
None confirmed in the sources
Work listed
2
Sources cited
7
Profile reviewed

Image and video editing

Valevski's papers, mostly written with Yaniv Leviathan and Yossi Matias, focus on controlling and editing the output of diffusion models. UniTune (2022) edits a single image from a text instruction by fine-tuning a diffusion model on that image, and Face0 (2023) conditions a text-to-image model on a face at sampling time, without fine-tuning or inversion. Valevski also co-authored Dreamix, which uses a video diffusion model for text-based editing of general videos, and a 2023 study showing that recaptioning training images improves how closely text-to-image models follow prompts.3456

Game engines and generated interfaces

In 2024 Valevski was co-first author of the GameNGen paper, which trains an agent to play DOOM and then a diffusion model that generates each next frame from past frames and player actions, so the game can be played in real time on a single TPU. The project page lists Valevski with Google Research. In 2026 Valevski co-authored "Generative UI: LLMs are Effective UI Generators", which argues that suitably prompted language models can generate whole interfaces rather than static text.217

Work and contributions

  • GameNGen (Diffusion Models Are Real-Time Game Engines) Published research

    Co-first author, with Yaniv Leviathan, Moab Arar, and Shlomi Fruchter, of the 2024 paper; the project page describes training a reinforcement-learning agent to play DOOM and then a diffusion model that generates the next frame conditioned on past frames and actions.21

  • UniTune Published research

    First author, with Matan Kalman, Eyal Molad, Eyal Segalis, Yossi Matias, and Yaniv Leviathan, of the 2022 paper on text-driven editing of an image by fine-tuning a diffusion model on that single image.3

Timeline

  1. Posted UniTune, on text-driven image editing with a fine-tuned diffusion model, as first author.3
  2. Co-authored Dreamix, a method for text-based editing of general videos with video diffusion models.5
  3. Posted Face0, on conditioning a text-to-image model on a face at sampling time, as first author.4
  4. Co-authored a study on recaptioning training data to improve text-to-image generation.6
  5. Posted GameNGen, a diffusion model that simulates DOOM in real time, as co-first author.21
  6. Co-authored "Generative UI: LLMs are Effective UI Generators".7

Documented roles

The sources do not confirm a current role; these are the roles they document.

  • Listed with Google Research as co-first author of GameNGen (2024), Google Research1

Companies in their history

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

7 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. GameNGen — Diffusion Models Are Real-Time Game Engines (project page)GameNGen authors · Official page · accessed Oct 21, 2026
  2. Diffusion Models Are Real-Time Game Engines (arXiv 2408.14837)arXiv · Paper · published Aug 27, 2024 · accessed Oct 21, 2026
  3. UniTune: Text-Driven Image Editing by Fine Tuning a Diffusion Model on a Single Image (arXiv 2210.09477)arXiv · Paper · published Oct 17, 2022 · accessed Oct 21, 2026
  4. Face0: Instantaneously Conditioning a Text-to-Image Model on a Face (arXiv 2306.06638)arXiv · Paper · published Jun 11, 2023 · accessed Oct 21, 2026
  5. Dreamix: Video Diffusion Models are General Video Editors (arXiv 2302.01329)arXiv · Paper · published Feb 2, 2023 · accessed Oct 21, 2026
  6. A Picture is Worth a Thousand Words: Principled Recaptioning Improves Image Generation (arXiv 2310.16656)arXiv · Paper · published Oct 25, 2023 · accessed Oct 21, 2026
  7. Generative UI: LLMs are Effective UI Generators (arXiv 2604.09577)arXiv · Paper · published Feb 24, 2026 · accessed Oct 21, 2026