MAViS is an end-to-end multi-agent collaborative framework for feature-length video storytelling. It coordinates specialized agents across multiple stages, including script writing, shot design, character modeling, keyframe generation, video animation, and audio generation. At each stage, the agents operate according to the 3E principle (Explore, Review, Enhance) to ensure the completeness of the intermediate output. Considering the functional limitations of current generative models, we propose script writing guidelines to optimize compatibility between scripts and generation tools. Experimental results demonstrate that MAViS achieves state-of-the-art performance in assistive features, visual quality, and video expressiveness. The modular framework further enhances extensibility with various generative models and tools. With simple user prompts, MAViS generates high-quality, expressive feature-length video storytelling, enriching users' inspiration and creativity. MAViS is the only framework that provides multimodal design outputs, such as videos with narrative and background music.
