SkyReels-V4: Multi-modal Video-Audio Generation, Inpainting and Editing model
Abstract
SkyReels V4 is a unified multi modal video foundation model for joint video audio generation, inpainting, and editing. The model adopts a dual stream Multimodal Diffusion Transformer (MMDiT) architecture, where one branch synthesizes video and the other generates temporally aligned audio, while sharing a powerful text encoder based on the Multimodal Large Language Models (MLLM). SkyReels V4 accepts rich multi modal instructions, including text, images, video clips, masks, and audio references. By combining the MLLMs multi modal instruction following capability with in context learning in the video branch MMDiT, the model can inject fine grained visual guidance under complex conditioning, while the audio branch MMDiT simultaneously leverages audio references to guide sound generation. On the video side, we adopt a channel concatenation formulation that unifies a wide range of inpainting style tasks, such as image to video, video extension, and video editing under a single interface, and naturally extends to vision referenced inpainting and editing via multi modal prompts. SkyReels V4 supports up to 1080p resolution, 32 FPS, and 15 second duration, enabling high fidelity, multi shot, cinema level video generation with synchronized audio. To make such high resolution, long-duration generation computationally feasible, we introduce an efficiency strategy: Joint generation of low resolution full sequences and high-resolution keyframes, followed by dedicated super-resolution and frame interpolation models. To our knowledge, SkyReels V4 is the first video foundation model that simultaneously supports multi-modal input, joint video audio generation, and a unified treatment of generation, inpainting, and editing, while maintaining strong efficiency and quality at cinematic resolutions and durations.
Keywords
Cite
@article{arxiv.2602.21818,
title = {SkyReels-V4: Multi-modal Video-Audio Generation, Inpainting and Editing model},
author = {Guibin Chen and Dixuan Lin and Jiangping Yang and Youqiang Zhang and Zhengcong Fei and Debang Li and Sheng Chen and Chaofeng Ao and Nuo Pang and Yiming Wang and Yikun Dou and Zheng Chen and Mingyuan Fan and Tuanhui Li and Mingshan Chang and Hao Zhang and Xiaopeng Sun and Jingtao Xu and Yuqiang Xie and Jiahua Wang and Zhiheng Xu and Weiming Xiong and Yuzhe Jin and Baoxuan Gu and Binjie Mao and Yunjie Yu and Jujie He and Yuhao Feng and Shiwen Tu and Chaojie Wang and Rui Yan and Wei Shen and Jingchen Wu and Peng Zhao and Xuanyue Zhong and Zhuangzhuang Liu and Kaifei Wang and Fuxiang Zhang and Weikai Xu and Wenyan Liu and Binglu Zhang and Yu Shen and Tianhui Xiong and Bin Peng and Liang Zeng and Xuchen Song and Haoxiang Guo and Peiyu Wang and Max W. Y. Lam and Chien-Hung Liu and Yahui Zhou},
journal= {arXiv preprint arXiv:2602.21818},
year = {2026}
}