Senior Research Scientist at Fish Audio.
I work on generative audio, voice LLMs, and multimodal foundation models.
With 7+ years of experience from research to production, I build product-ready speech and audio AI systems across expressive TTS, full-duplex speech systems, codec and tokenization design, large-scale multi-GPU training, and deployment.
- Website: zengchang233.github.io
- Google Scholar: Profile
- GitHub: @zengchang233
- LinkedIn: chang-zeng-5451a4191
- X (Twitter): @zzzzzc12
- Generative Audio and Voice LLMs
- Multimodal Foundation Models
- Speech and Singing Voice Generation
- Speaker Recognition and Antispoofing
- Audio Separation and Enhancement
2026.08 - present: Senior Research Scientist, Fish Audio2026.05 - 2026.08: Senior Research Scientist, StepFun2025.09 - 2026.05: Senior AI Researcher, Shanda AI Research Tokyo2024.04 - 2025.09: Multimodal Generative AI Researcher, Li Auto
2026.10: Released Uni-Stream on arXiv.2026.10: Released Fish Audio Drama 3 Technical Report on arXiv.2026.10: Released Can LLM Agents Automate Reinforcement Learning for Text-to-Speech? on arXiv.2026.09: Released UNITE-AUDIO on arXiv.2026.06: Speech Codec Probing from Semantic and Phonetic Perspectives accepted by Interspeech 2026.2026.05: Released StepAudio 2.5 Technical Report on arXiv.2026.05: A Semantically Consistent Dataset for Data-Efficient Query-Based Universal Sound Separation accepted by ICML 2026.2026.03: Released BEAVER on arXiv.2026.01: DrivingScene, PAGS, and Automatic DER Estimation accepted by ICASSP 2026.2025.12: Released Towards Interactive Intelligence for Digital Humans on arXiv.
- Uni-Stream: A Unified Streaming Speech Language Model for Speech Synthesis and Recognition (arXiv, 2026)
- Fish Audio Drama 3 Technical Report (arXiv, 2026)
- Can LLM Agents Automate Reinforcement Learning for Text-to-Speech? (arXiv, 2026)
- UNITE-AUDIO: Joint Learning of Continuous Tokenization and Latent Flow Matching for Text-to-Audio Generation (arXiv, 2026)
- Speech Codec Probing from Semantic and Phonetic Perspectives (Interspeech 2026)
- StepAudio 2.5 Technical Report (arXiv, 2026)
- Automatic Estimation of Speaker Diarization Error Rate Based on Features of Audio Quality and Speaker Discriminability (ICASSP 2026)
- Attention Back-end for Automatic Speaker Verification with Multiple Enrollment Utterances (ICASSP 2022)
- Xiaoicesing 2: A High-Fidelity Singing Voice Synthesizer Based on Generative Adversarial Network (Interspeech 2023)
- HiFi-WaveGAN: Generative Adversarial Network with Auxiliary Spectrogram-Phase Loss for High-Fidelity Singing Voice Generation (ISNN 2024)