Yuxiang Zhao

Email: yuxiangzhao@sjtu.edu.cn
Research Interests: speech translation, speech security

Education

Shanghai Jiao Tong University

M.S. Student, School of Computer Science | Sep. 2024 - Present

Lab: X-LANCE Lab
Advisor: Prof. Xie Chen

Shanghai Jiao Tong University

B.Eng. in Civil Engineering | Sep. 2020 - Jun. 2024

Research Experience

Speech Translation and Low-Latency Simultaneous Interpretation

Oct. 2025 - Present

Built X-Translator, a cascaded ASR+MT+TTS speech-to-speech translation system for real-time spoken interaction, achieving 3-5 seconds of end-to-end latency with voice cloning support.
Designed a multidimensional evaluation framework for speech translation, covering speech-to-text translation, speech-to-speech translation, offline generation, and streaming generation.
Evaluated translation quality, speech quality, speaker preservation, emotion and paralinguistic fidelity, temporal consistency, and streaming latency in a unified protocol.
Studied machine translation training without target-language data by using LLM-generated synthetic target-side data to train compact NMT models, analyzing the effects of distribution coverage, sample quality, and sequence-level supervision.

Speech Security, Anonymization, and Traceable Generation

Oct. 2024 - Oct. 2025

Proposed a watermark-free TTS traceability framework that jointly trains a generation model and discriminator for model-level attribution while preserving speech quality.
Improved traceability robustness and generalization in cross-domain scenarios without relying on explicit speech watermarks.
Built a utility-preserved speech anonymization framework that combines flow-matching-based speaker anonymization with generative content editing.
Reduced identity and content leakage risks while preserving the usefulness of anonymized data for downstream ASR, TTS, and other speech tasks.

Publications

Yuxiang Zhao, Yunchong Xiao, Yushen Chen, Zhikang Niu, Shuai Wang, Kai Yu, Xie Chen^†.
Traceable TTS: Toward Watermark-Free TTS with Strong Traceability.
Submitted to INTERSPEECH 2026.
[Paper] [Code]
Yuxiang Zhao, Yanjie An, Hanzheng Li, Xuhui Shen, Keqi Deng, Shuai Fan, Kai Yu, Xie Chen^†.
Distilling Compact Machine Translation Models from Synthetic Target-Side Data.
Submitted to EMNLP 2026.
Yunchong Xiao^*, Yuxiang Zhao^*, Ziyang Ma, Shuai Wang, Kai Yu, Jiachun Liao, Xie Chen^†.
Anonymization, Not Elimination: Utility-Preserved Speech Anonymization.
IEEE Transactions on Audio, Speech and Language Processing, 2026.
[Paper]
Yanjie An^*, Yuxiang Zhao^*, Yichi Zhang, Qixi Zheng, Yujie Tu, Keqi Deng, Kai Yu, Xie Chen^†.
OpenSTBench: Beyond Semantic Evaluation for Speech Translation.
Submitted to EMNLP 2026.
[Paper] [Code]
Qixi Zheng, Yuxiang Zhao, Tianrui Wang, Wenxi Chen, Kele Xu, Yikang Li, Qinyuan Chen, Xipeng Qiu, Kai Yu, Xie Chen^†.
X-VC: Zero-shot Streaming Voice Conversion in Codec Space.
Submitted to ACM MM 2026.
[Paper] [Code] [Demo]
Ruiqi Yan, Wenxi Chen, Zhanxun Liu, Ziyang Ma, Haopeng Lin, Hanlin Wen, Hanke Xie, Jun Wu, Yuzhe Liang, Yuxiang Zhao, Pengchao Feng, Jiale Qian, Hao Meng, Yuhang Dai, Shunshun Yin, Ming Tao, Lei Xie, Kai Yu, Xinsheng Wang, Xie Chen^†.
SoulX-Duplug: Plug-and-Play Streaming State Prediction Module for Realtime Full-Duplex Speech Conversation.
Submitted to INTERSPEECH 2026.
[Paper] [Code] [Model] [Dataset]

^* Equal contribution. ^† Corresponding author.

Internship

AISpeech Co., Ltd.

Speech Research Intern | Jul. 2025 - Present