Reproduce “AutoMV: An Automatic Multi-Agent System for Music Video Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiaoxuan Tang, Xinping Lei, Chaoran Zhu, Shiyun Chen, Ruibin Yuan, Yizhi Li, Changjae Oh, Ge Zhang and 5 more
AutoMV is a training-free, multi-agent system that autonomously creates coherent, full-length music videos directly from a song by integrating music informatics with LLM-based planning and verification. The system from Beijing University of Posts and Telecommunications and collaborators significantly outperforms commercial baselines in expert evaluations for technical quality, content alignment, and artistic metrics, while offering an accessible content creation solution.
Key insights
- Multi-agent coordination driven by comprehensive music-aware reasoning can successfully tackle complex, long-form cross-modal generation challenges.
- Large Language Models (LLMs) demonstrate capability as orchestrators for high-level planning and as effective iterative verifiers for quality control in generative pipelines.
- Maintaining character identity across diverse scenes requires dedicated character banks, while lyric-driven narrative planning enhances music-visual alignment and storytelling.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.