[AAAI'26] Bring Your Dreams to Life: Continual Text-to-Video Customization
December 9, 2025 路 View on GitHub
Jiahua Dong, Xudong Wang, Wenqi Liang, Zongyan Han, Meng Cao, Duzhen Zhang, Hanbin Zhao, Zhi Han, Salman Khan, and Fahad Shahbaz Khan, "Bring Your Dreams to Life: Continual Text-to-Video Customization", AAAI, 2026. [arXiv]
馃敟馃敟馃敟 News
- 2025-12-09: Codes on both DreamVideo and Wan 2.1 are released. 猸愶笍猸愶笍猸愶笍
- 2025-12-01: This repo is released.
- 2025-11-08: CCVD is accepted at AAAI 2026. 馃帀馃帀馃帀
Abstract: Customized text-to-video generation (CTVG) has recently witnessed significant progress in generating tailored videos from user-specific text. However, existing CTVG methods unrealistically assume that personalized concepts remain static and do not expand incrementally over time. Additionally, they struggle with catastrophic forgetting and concept neglect when continuously learning new concepts, including subjects and motions. To resolve the above challenges, we develop a novel Continual Customized Video Diffusion (CCVD) model, which can continuously learn new concepts to generate videos across various text-to-video generation tasks by tackling catastrophic forgetting and concept neglect. Specifically, to address catastrophic forgetting, we introduce a concept-specific attribute retention module and a task-aware concept aggregation strategy. They can capture the unique characteristics and identities of old concepts during training, while combining all subject and motion adapters of old concepts based on their relevance during testing. Furthermore, to tackle concept neglect, we develop a controllable conditional synthesis to enhance regional features and align video contexts with user conditions, by incorporating layer-specific region attention-guided noise estimation. Extensive experimental comparisons demonstrate that our CCVD outperforms existing CTVG baselines on both the DreamVideo and Wan 2.1 backbones.
The architectural overview of our CCVD model. It includes (a) a concept-specific attribute retention module, (b) a task-aware concept aggregation strategy to overcome catastrophic forgetting of previous concepts during training and testing, and (c) a controllable conditional synthesis module with layer-specific region attention and attention-guided noise estimation to address the issue of concept neglect.
Continual Text-to-Video Customization Task
Experiment Task Setting
Examples
Multi-concept Results
路 DreamVideo Baseline
路 Wan Baseline
| Base: The desert with blue sky in the background. Region: V31 cat and V12 dog walking in the desert | Base: A river with flowers and plants on its banks. Region: V22 duck toy and V32 dog playing on the river. | Base: A street聽with buildings. Region: V22 dog and V6 bear toy and V31 cat walking in the street. | |
|---|---|---|---|
| Ours | |||
| L2DM | |||
| CLoRA | |||
| LoRA-M |
Single-concept Results
路 DreamVideo Baseline
路 Wan Baseline
| V17 man wearing a chef's hat, cutting vegetables in the kitchen. | V1 dog running on the park lawn. | V31 cat wearing sunglasses lying on a beach chair on a sunset beach. | |
|---|---|---|---|
| Ours | |||
| L2DM | |||
| CLoRA | |||
| LoRA-M |
Style Transfer
Video Editing
Implementation
馃敟 If you wish to build upon the DreamVideo baseline model, please refer to DreamVideo.
馃敟 If you wish to build upon the Wan baseline model, please refer to Wan.
Acknowledgement
This project is mainly based on DreamVideo and Wan. In our experiments, we use the following projects:
[1] DreamVideo: Composing Your Dream Videos with Customized Subject and Motion.
[2] Wan: Open and Advanced Large-Scale Video Generative Models.
[3] How to Continually Adapt Text-to-Image Diffusion Models for Flexible Customization?
Contact
If you have any questions, you are very welcome to email dongjiahua1995@gmail.com or wangxudong@sia.cn.
BibTeX
If you find CCVD useful for your research and applications, please cite our paper using this BibTeX:
@inproceedings{AAAI2026_CCVD_Bring,
title={Bring Your Dreams to Life: Continual Text-to-Video Customization},
author={Jiahua Dong and Xudong Wang and Wenqi Liang and Zongyan Han and Meng Cao and Duzhen Zhang and Hanbin Zhao and Zhi Han and Salman Khan and Fahad Shahbaz Khan},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
year={2025}
}