Versatile Framework for Song Generation with Prompt-based Control
Yu Zhang, Wenxiang Guo, Changhao Pan, Zhiyuan Zhu, Ruiqi Li, Jingyu Lu, Rongjie Huang, Ruiyuan Zhang, Zhiqing Hong, Ziyue Jiang, Zhou Zhao
2025-04-29
Summary
This paper talks about VersBand, a new AI system that can create complete songs, including both singing and background music, all based on what the user asks for in a prompt.
What's the problem?
The problem is that most song-making AIs either focus only on generating music or just vocals, and they usually can't make both parts work together smoothly or let users control the style and details easily.
What's the solution?
The researchers built VersBand by combining several different AI models so it can handle both singing and music at the same time. They also made it so users can guide the song creation process by giving prompts, like describing the mood, style, or even some lyrics, which helps the AI make songs that match what the user wants.
Why it matters?
This matters because it opens up new creative possibilities for musicians, songwriters, and even people with no music experience, making it much easier to create custom songs for fun, education, or professional use.
Abstract
VersBand is a multi-task song generation framework that integrates multiple models for high-quality, aligned vocals and accompaniments with prompt-based control.