WanSong v1.0 Technical Report

2026-07-16

AI Summary

Generating high-fidelity, controllable, long-form multilingual songs remains a significant challenge. This work proposes WanSong, an end-to-end pure diffusion model that eschews autoregressive or cascaded architectures and instead generates up to five-minute dual-track (vocals and accompaniment) multilingual songs in a single pass. By integrating step distillation to accelerate inference, WanSong supports flexible conditional control and efficient fine-tuning, substantially enhancing both generation efficiency and editability. Experimental results demonstrate that WanSong produces long audio clips of commercial-grade audio quality, achieving rapid inference and effective downstream customization while preserving high fidelity.

Abstract

Music generation foundation models have recently attracted significant industry attention. However, achieving efficient generation and high-fidelity long-form audio while supporting controllability remains challenging. To address these needs, we present \textbf{WanSong}, a simple yet powerful approach for long-form, commercial-grade song generation. Unlike autoregressive (AR) and cascaded multi-stage pipelines (\eg, AR followed by diffusion), \textbf{WanSong} is a pure diffusion-based model that directly generates high-fidelity, multilingual songs up to 5 minutes and outputs dual stems (vocals and background music) in a single run. In addition, our diffusion framework enables faster inference through step-distillation, and offers an efficient pathway for fine-tuning and customization to support downstream editing tasks.

Problem

• music generation
• long-form audio
• controllability
• high-fidelity
• dual stems

Innovation

• diffusion-based model
• long-form music generation
• dual-stem output
• step-distillation
• controllable audio synthesis