Qwen-Music Technical Report

Hugging Face Daily Papers Papers
music-generation tokenizer chain-of-thought audio-generation deep-learning qwen

Summary

Qwen-Music is a music generation model that produces high-fidelity songs with vocals, supporting text-to-music and cover song generation. It uses a novel Melody-Chain-of-Thought mechanism and achieves state-of-the-art results on 13 of 16 objective metrics.

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.
Original Article
View Cached Full Text

Cached at: 07/20/26, 09:39 AM

Paper page - Qwen-Music Technical Report

Source: https://huggingface.co/papers/2607.11699 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

Inthisreport,weintroduceQwen-Music,apowerfulmusicgenerationmodelcapableofproducinghighlymusicalandhigh-fidelitysongswithcompletevocalsinging.Qwen-Musicsupportstwocoretasks:TexttoMusicGeneration,whichcreateentirelynewsongsfromtextdescriptions,lyrics,andmusicalattributes,andCoverSongGeneration,whichreinterpretsexistingsongswithdifferentstylesandvocalcharacteristics.Architecturally,Qwen-Musicintegratesthreecorecomponents:Qwen-Music-Tokenizer,Qwen-Music-LLM,andQwen-Music-Render.Qwen-Music-Tokenizercompressesaudiointoa25Hzsingle-codebookstreamofMusicSemanticTokensthatpreservesemanticandmelodicinformationforLLMprediction.Basedonthesetokens,Qwen-Music-LLMperformsautoregressivemusicsemanticmodeling,withakeynoveltybeingamelody-token-basedchain-of-thought(Melody-CoT)mechanismthatplansmelodiesbeforefull-songgeneration,improvingcreativity,musicality,structuralcoherence,andreference-audio-basedmelodycloning.Toovercomethefidelitylimitationsofdiscretesemantictokens,Qwen-Music-Renderperformsgenerativestereorendering,enrichingacousticdetailsandproducinghigh-fidelitystereowaveforms.Finally,wetrainQwen-Music-LLMonmorethan5millionhoursofmultilingualmusicdatacoveringhundredsoflanguages.Wefirstapplyquality-awarepre-trainingcurriculum,thenuseprogressivepost-training,comprisingsupervisedinitialization,offlineDPO,andonlineGSPO,tofurtherimprovemusicalityandinstruction-followingability.Across600ChineseandEnglishprompts,Qwen-Musicachievesstate-of-the-artresultsin13of16objectivemusicalityandaudio-qualitymetrics.ProfessionalevaluatorsalsopreferQwen-Musicoverleadingproprietarysystems.Forcoversonggeneration,Qwen-Musicpreservesreferencemelodiesmoreaccuratelythanleadingproprietarysystems.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.11699

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.11699 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.11699 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.11699 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Qwen-Image-2.0 Technical Report

Hugging Face Daily Papers

Qwen-Image-2.0 is a new image generation foundation model that unifies high-fidelity synthesis and precise editing using Qwen3-VL and a Multimodal Diffusion Transformer. It excels in text-rich content, multilingual typography, and photorealistic generation.

Qwen3-TTS Technical Report

Papers with Code Trending

The Qwen3-TTS technical report introduces a series of advanced multilingual text-to-speech models with voice cloning and controllable generation, featuring a dual-track LM architecture and specialized tokenizers for low-latency streaming.

Qwen3.5-Omni Technical Report

Hugging Face Daily Papers

Qwen3.5-Omni is a hundreds-of-billions-parameter multimodal model with advanced audio-visual understanding and generation capabilities, featuring novel Audio-Visual Vibe Coding and achieving SOTA results across 215 benchmarks while matching Gemini-3.1 Pro.