FreeMem: Enhancing Consistency in Long Video Generation via Tuning-Free Memory
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (10) · 8340-8348
DOI: 10.1609/aaai.v40i10.37783 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Jibin Peng und 9 weiteren Autor:innen.
Abstract (Originalsprache)
Text-to-Video (T2V) generation has advanced greatly, yet maintaining consistency remains challenging, especially for tuning-free long video generation. We attribute the consistency problem to cumulative deviations for long video generation at three levels: the random noise lacking correlation results initial deviation between frames; discrepancy in semantic feature tokens between denoising network blocks gradually accumulates as the frame count grows, leading to greater deviations; attention mechanisms struggle to capture global relationships across distant frames in long videos. To address these, we propose FreeMem, a tuning-free framework leveraging hierarchical memory update and injection: the noise memory stabilizes consistency by manipulating low and high frequency components in the initial noise space; the token memory combats inconsistency through adaptive fusion of historical and current semantic feature tokens between denoising network blocks; and the attention memory establishes persistent cache to model long-range relationships within self attention layers. Evaluated on VBench, FreeMem improves subject and background consistency matrics across various methods, offering a practical solution for low-cost, high-consistency long video generation.
Zitieren
APA
Peng, J., Lin, D., Xu, Z., Lu, H., Liu, R., Xie, W., Wang, M., Liang, L., Wang, Y., & Guo, Q. (2026). FreeMem: Enhancing Consistency in Long Video Generation via Tuning-Free Memory. Proceedings of the AAAI Conference on Artificial Intelligence, 40(10), 8340-8348. https://doi.org/10.1609/aaai.v40i10.37783
Vancouver
Peng J, Lin D, Xu Z, Lu H, Liu R, Xie W, et al. FreeMem: Enhancing Consistency in Long Video Generation via Tuning-Free Memory. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(10):8340-8348. doi:10.1609/aaai.v40i10.37783
BibTeX
@article{Peng2026,
author = {Peng, Jibin and Lin, Di and Xu, Zhecheng and Lu, Haoran and Liu, Ruonan and Xie, Wuyuan and Wang, Miaohui and Liang, Lingyu and Wang, Yi and Guo, Qing},
title = {FreeMem: Enhancing Consistency in Long Video Generation via Tuning-Free Memory},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {10},
pages = {8340--8348},
doi = {10.1609/aaai.v40i10.37783}
}
EUVIMED