(Hover over to view inputs)

Longer Video Inference (32 frames)


Creating a vivid video from the event or scenario in our imagination is a truly fascinating experience. Recent advancements in text-to-video synthesis have unveiled the potential to achieve this with prompts only. While text is convenient in conveying the overall scene context, it may be insufficient to control precisely. In this paper, we explore customized video generation by utilizing text as context description and motion structure (e.g. frame-wise depth) as concrete guidance. Our method, dubbed Make-Your-Video, involves joint-conditional video generation using a Latent Diffusion Model that is pre-trained for still image synthesis and then promoted for video generation with the introduction of temporal modules. This two-stage learning scheme not only reduces the computing resources required, but also improves the performance by transferring the rich concepts available in image datasets solely into video generation. Moreover, we use a simple yet effective causal attention mask strategy to enable longer video synthesis, which mitigates the potential quality degradation effectively. Experimental results show the superiority of our method over existing baselines, particularly in terms of temporal coherence and fidelity to users' guidance. In addition, our model enables several intriguing applications that demonstrate potential for practical usage.

Method Overview


Real-life scene to video

Real-life scene Text2Video-zero+CtrlNet LVDMExt+Adapter Ours

"A dam discharging water"

"A futuristic rocket ship on a launchpad, with sleek design, glowing lights"

3D scene modeling to video

3D scene modeling Text2Video-zero+CtrlNet LVDMExt+Adapter Ours

"A train on the rail, 2D cartoon style"

"A Van Gogh style painting on drawing board in park, some books on the picnic blanket, photorealistic"

"A Chinese ink wash landscape painting"

Video re-rendering

Original video SD-Depth Text2Video-zero+CtrlNet LVDMExt+Adapter Tune-A-Video Ours

"A tiger walks in the forest, photorealistic"

"An origami boat moving on the sea"

"A filled chocolate moving on the road"

"A camel walking on the snow field, Miyazaki Hayao anime style"

"A waterfall in the middle of a glacier"

"An astronaut is walking on the moon, cartoon style"

Video re-rendering (with concurrent works)

Original video SD-Depth ControlVideo VideoComposer Gen-1 Ours

"A camel walking on the snow field, Miyazaki Hayao anime style"

"A toy cat sitting on the ground at Times Square, photorealistic"

"A black-and-white robot cow walking in a river"

"A waterfall in the middle of a glacier"


					  author  = {Xing, Jinbo and Xia, Menghan and Liu, Yuxin and Zhang, Yuechen and Zhang, Yong and He, Yingqing and Liu, Hanyuan and Chen, Haoxin and Cun, Xiaodong and Wang, Xintao and Shan, Ying and Wong, Tien-Tsin},
					  title   = {Make-Your-Video: Customized Video Generation Using Textual and Structural Guidance},
					  journal = {arXiv preprint arXiv:2306.00943},
					  year    = {2023}

This page is modified based on the LVDM project page.