LeVo – An AI singing model developed by Tencent AI Lab

What is LeVo?

LeVo is an AI singing model developed by Tencent AI Lab, known for its powerful voice cloning capabilities. With just a 3-second audio clip, LeVo can accurately replicate a target voice’s tone, emotion, and rhythm—without the need for extensive training data. It supports multi-track generation, enabling the separate creation of vocal and accompaniment tracks, which offers greater flexibility for post-production editing.

LeVo’s technical architecture is built on a language model (LM) framework, combining LeLM and a music codec. This allows it to generate high-quality audio tracks in parallel, with audio fidelity rivaling industry leaders and exceptional performance in lyric alignment.

Key Features of LeVo

• Zero-Shot Voice Cloning
  Replicates a target voice—including pitch, emotion, and rhythm—with only a 3-second audio sample, requiring no large-scale training data.

• Multi-Track Generation
  Supports dual-track output, generating vocals and accompaniment separately, which enhances flexibility in mixing and editing.

• High-Fidelity Musical Output
  LeVo delivers sound quality comparable to leading models in the industry. It excels in musicality, harmony between vocals and background music, and Mean Opinion Score (MOS). Generation results are optimized through multi-preference alignment to maintain high fidelity across various styles and scenarios.

Technical Architecture of LeVo

• Language Model-Based Design
  LeVo is based on a language model (LM) structure, integrating LeLM and a music codec that enables parallel generation of high-quality music content.

Performance of LeVo

• On Par with Industry Leaders
  LeVo achieves competitive performance across multiple key metrics compared to leading models like Suno 4.5.

• Superior Lyric Alignment
  LeVo surpasses Suno 4.5 in Lyric Alignment Capability (LYC) by 0.21 points, demonstrating strong text-to-music control and synchronization.

Project Link

• Official Website: https://levo-demo.github.io/

Application Scenarios

• Individual Music Creators
  Offers a low-barrier, high-quality platform for music lovers who may lack professional production skills.

• Professional Music Producers
  LeVo’s multi-track generation and high-fidelity output boost productivity and creative possibilities in professional workflows.

• Music Education Institutions
  Provides an engaging and innovative tool for music teaching and learning experiences.