Qwen-Music 技術報告

摘要

在本报告中，我们介绍Qwen-Music——一款强大的音乐生成模型，能够生成具有高度音乐性与高保真度、并包含完整人声演唱的歌曲。Qwen-Music支持两项核心任务：文字转音乐生成（Text to Music Generation），即根据文字描述、歌词及音乐属性创作全新歌曲；以及翻唱生成（Cover Song Generation），即用不同的风格和声音特征重新演绎现有歌曲。在架构上，Qwen-Music整合了三个核心组件：Qwen-Music-Tokenizer、Qwen-Music-LLM 和 Qwen-Music-Render。Qwen-Music-Tokenizer将音频压缩为25 Hz单码本流的音乐语义标记（Music Semantic Tokens），这些标记保留了语义与旋律信息，供大语言模型预测。基于这些标记，Qwen-Music-LLM进行自回归音乐语义建模，其关键创新在于一种基于旋律标记的链式推理（Melody-CoT）机制：在生成整首歌曲之前先规划旋律，从而提升创造性、音乐性、结构连贯性，并支持基于参考音频的旋律克隆。为克服离散语义标记在保真度上的局限，Qwen-Music-Render执行生成式立体声渲染，丰富声学细节，产生高保真立体声波形。最后，我们在超过500万小时的多语言音乐数据（涵盖数百种语言）上训练了Qwen-Music-LLM。我们首先应用质量感知的预训练课程，然后采用渐进式后训练——包括监督初始化、离线直接偏好优化（DPO）和在线生成式自我偏好优化（GSPO）——以进一步提升音乐性和指令遵循能力。在600条中英文提示的评估中，Qwen-Music在16项客观音乐性与音频质量指标中，有13项达到最先进水平。专业评估员也更倾向于选择Qwen-Music而非领先的专有系统。在翻唱生成任务中，Qwen-Music比领先的专有系统更准确地保留参考旋律。

English

In this report, we introduce Qwen-Music, a powerful music generation model capable of producing highly musical and high-fidelity songs with complete vocal singing. Qwen-Music supports two core tasks: Text to Music Generation, which create entirely new songs from text descriptions, lyrics, and musical attributes, and Cover Song Generation, which reinterprets existing songs with different styles and vocal characteristics. Architecturally, Qwen-Music integrates three core components: Qwen-Music-Tokenizer, Qwen-Music-LLM, and Qwen-Music-Render. Qwen-Music-Tokenizer compresses audio into a 25 Hz single-codebook stream of Music Semantic Tokens that preserve semantic and melodic information for LLM prediction. Based on these tokens, Qwen-Music-LLM performs autoregressive music semantic modeling, with a key novelty being a melody-token-based chain-of-thought (Melody-CoT) mechanism that plans melodies before full-song generation, improving creativity, musicality, structural coherence, and reference-audio-based melody cloning. To overcome the fidelity limitations of discrete semantic tokens, Qwen-Music-Render performs generative stereo rendering, enriching acoustic details and producing high-fidelity stereo waveforms. Finally, we train Qwen-Music-LLM on more than 5 million hours of multilingual music data covering hundreds of languages. We first apply quality-aware pre-training curriculum, then use progressive post-training, comprising supervised initialization, offline DPO, and online GSPO, to further improve musicality and instruction-following ability. Across 600 Chinese and English prompts, Qwen-Music achieves state-of-the-art results in 13 of 16 objective musicality and audio-quality metrics. Professional evaluators also prefer Qwen-Music over leading proprietary systems. For cover song generation, Qwen-Music preserves reference melodies more accurately than leading proprietary systems.