Hunyuan-GameCraft: High-dynamic Interactive Game Video Generation with Hybrid History Condition
Jiaqi Li, Junshu Tang, Zhiyong Xu, Longhuang Wu, Yuan Zhou, Shuai Shao, Tianbao Yu, Zhiguo Cao, Qinglin Lu · 2025
Read Hunyuan-GameCraft: High-dynamic Interactive Game Video Generation with Hybrid History Condition by Jiaqi Li, Junshu Tang, Zhiyong Xu, Longhuang Wu, Yuan Zhou, Shuai Shao, Tianbao Yu, Zhiguo Cao, Qinglin Lu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-17201.txt.
Description
Jiaqi Li and co-authors present a system that makes playable game video that reacts to the player, in work from twenty twenty-five. Recent video models can make smooth, good-looking clips from text or images. Game use needs more: fast action, long play, and tight control. Past methods fall short on pace, range, and low cost. The new system aims to close all four gaps at once.
Keyboard and mouse inputs are mapped into one shared camera space. That mapping lets moves and views blend without jarring jumps. A mixed history training plan grows clips step by step while keeping the scene in mind. A slimmed-down model then cuts the compute cost for live play. Training used over a million clips from more than a hundred top games. A clean, hand-tagged set then sharpened control and looks. Tests show clear gains over past models.
Catalog record
| CREATOR | Jiaqi Li, Junshu Tang, Zhiyong Xu, Longhuang Wu, Yuan Zhou, Shuai Shao, Tianbao Yu, Zhiguo Cao, Qinglin Lu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-17201