Dongchao Yang

About

I am a PhD student at The Chinese University of Hong Kong, majoring in Speech and Audio Processing, supervised by Prof. Helen Meng. Before that, I received the Master's Degree from Peking University in 2023.

My research focuses on developing a human-agent that can communicate with humans, e.g. understanding human's speech and environment sounds, and then producing feedback to humans.

Research Interests

News

• 2025 ALMTokenizer accepted by ICML 2025!
• 2025 SimpleSpeech 2 accepted by IEEE TASLP 2025.
• 2025 Recognized as ICLR Notable Reviewer and NeurIPS Top Reviewer.
• 2024 Received IEEE SPS Young Author Best Paper Award for Diffsound!
• 2024 Won ISCA Best Student Paper Award at Interspeech 2024 for SimpleSpeech.
• 2024 UniAudio highlighted by Stanford AI Index Report 2024 as one of only three audio papers (alongside Google's MusicLM and Meta's MusicGen).
• 2024 UniAudio accepted by ICML 2024; UniAudio 1.5 accepted by NeurIPS 2024.

Education

• The Chinese University of Hong Kong
  PhD Student, SEEM
  Aug 2023 - Present
• Peking University
  Master's Degree, School of Electronic and Computer Engineering
  Aug 2020 - Aug 2023
• Shanghai University
  Bachelor's Degree, School of Computer Engineering and Science
  Aug 2016 - Jul 2020

Experience

• Microsoft Research Asia (MSRA)
  Research Intern, Speech Group · Supervisor: Xu Tan
  Jul 2023 - Sep 2023
• Tencent AI Lab
  Research Intern, Speech Group · Supervisors: Songxiang Liu, Chao Weng, Bo Wu
  May 2021 - May 2023

Highlighted Research

ICML 2024 Stanford AI Index Report 2024

UniAudio: An Audio Foundation Model Toward Universal Audio Generation

A unified audio foundation model capable of generating speech, music, sound effects, and more within a single framework. One of only three audio papers highlighted in the Stanford AI Index Report 2024, alongside Google's MusicLM and Meta's MusicGen.

Dongchao Yang*, Jinchuan Tian*, Xu Tan, Rongjie Huang, et al.

Paper Code AI Index Report

IEEE TASLP 2023 IEEE SPS Young Author Best Paper Award

Diffsound: Discrete Diffusion Model for Text-to-sound Generation

The first text-to-audio generation work. Proposes a discrete diffusion model to generate diverse and high-quality sound effects directly from text descriptions, opening up the text-to-audio generation research direction.

Dongchao Yang, Jianwei Yu, Helin Wang, Wen Wang, Chao Weng, Yuexian Zou, Dong Yu

Paper Code

Interspeech 2024 ISCA Best Student Paper Award

SimpleSpeech: Towards Simple and Efficient Text-to-Speech with Scalar Latent Transformer Diffusion Models

The first non-autoregressive TTS that supports large-scale speech training without frame-level annotations. Uses scalar latent transformer diffusion models to achieve high-quality speech synthesis with significantly reduced complexity and inference cost.

Dongchao Yang, Dingdong Wang, Haohan Guo, Xueyuan Chen, Xixin Wu, Helen Meng

Paper Code

IEEE TASLP 2024

InstructTTS: Modelling Expressive TTS in Discrete Latent Space with Natural Language Style Prompt

The first Chinese TTS system that supports natural language prompt-controlled speech style generation. Enables intuitive and flexible control over speech style, emotion, and prosody through simple text instructions.

Dongchao Yang*, Songxiang Liu*, Rongjie Huang, Chao Weng, Helen Meng

Paper Code

NeurIPS 2024

UniAudio 1.5: Large Language Model-driven Audio Codec is A Few-shot Audio Task Learner

Demonstrates that an LLM-driven audio codec can serve as a powerful few-shot learner across diverse audio tasks, unifying audio understanding and generation through a codec-based approach.

Dongchao Yang, Haohan Guo, Yuanyuan Wang, Rongjie Huang, Xiang Li, Xu Tan, Xixin Wu, Helen Meng

Paper Code

ICML 2025

ALMTokenizer: A Low-bitrate and Semantic-rich Audio Codec Tokenizer for Audio Language Modeling

Proposes a new codec paradigm that achieves extremely low bitrate while preserving rich semantic information, specifically designed for audio language modeling and enabling more efficient audio-LLM integration.

Dongchao Yang, Songxiang Liu, Haohan Guo, Jiankun Zhao, Yuanyuan Wang, et al.

Paper

All Publications

* denotes equal contributions.

Services

• IEEE/CAA Journal of Automatica Sinica Reviewer
• ICASSP, InterSpeech Reviewer
• ICML, NeurIPS, ICLR, ACM-MM, COLM, IJCAI Reviewer
• IEEE Transactions on Audio, Speech, and Language Processing Reviewer
• IEEE Signal Processing Letters Reviewer

Selected Awards

• ICLR Notable Reviewer 2025
• NeurIPS Top Reviewer 2025
• IEEE SPS Young Author Best Paper Award 2024
• ISCA Best Student Paper Award 2024
• Outstanding Graduate of Peking University 2023
• Excellent Graduation Thesis of Peking University 2023
• 1st Team Ranking of DCASE Challenge Task 5 (Judges' Award) 2021
• Bronze Prize of ACM-ICPC Asia Regional Competition 2018