# Diffusion Models Are Real-Time Game Engines

*Source* · `src-valevski-2024-gamengen` · status: **n/a**

| field | value |
|---|---|
| kind | Peer-reviewed paper |
| authors | Dani Valevski, Yaniv Leviathan, Moab Arar, Shlomi Fruchter |
| venue | ICLR 2025 |
| date | 2024-08 |
| url | https://arxiv.org/abs/2408.14837 |
| arxiv | 2408.14837 |
| accessed | 2026-10-24 |

## Notes

GameNGen: DOOM at 20 fps on one TPU-v5; PSNR 29.43; raters 58%/60% on 1.6 s/3.2 s clips; 64-frame context (~3 s); trained on 70M frames from a PPO agent (50M env steps); 128 TPU-v5e for training; limitations: memory, heuristic errors, no game creation.

## Backlinks (computed)

**cited by:** [Diffusion and autoregressive neural game engines](../technique/diffusion-game-engines.md)

## Citation

Dani Valevski, Yaniv Leviathan, Moab Arar, Shlomi Fruchter (2024-08). *Diffusion Models Are Real-Time Game Engines*. ICLR 2025 https://arxiv.org/abs/2408.14837

---
Provenance: created in run `2026-10-24-techniques-world-models`.