Scholar
Songxiang Liu
Google Scholar ID: 4fD1l28AAAAJ
Meituan multi-modal team, PhD (The Chinese University of Hong Kong)
Citations & Impact
Citations
2,066
H-index
23
i10-index
38
Publications
20
Co-authors
0
Publications
Resume
Academic Achievements
- 2025: 'ALMTokenizer: A Low-bitrate and Semantic-rich Audio Codec Tokenizer for Audio Language Modeling' accepted by ICML 2025
- Apr 2025: Released technical report of Kimi-Audio (with code, model, and paper)
- 2024: 'UniAudio: Towards Universal Audio Generation with Large Language Models' accepted by ICML 2024
- 2024: 'InstructTTS: Modelling Expressive TTS in Discrete Latent Space with Natural Language Style Prompt' accepted by IEEE/ACM TASLP (corresponding author)
- 2023: Released InstructTTS on expressive TTS with natural language prompts (arXiv)
- 2022: Published DiffGAN-TTS paper (arXiv)
- 2021: DiffSVC paper accepted by ASRU 2021
- 2021: Published singing voice conversion work using denoising diffusion probabilistic models (DDPM) (arXiv)
- 2021: FastSVC paper accepted as an oral presentation at ICME 2021
- Multiple IEEE/ACM TASLP journal papers on voice conversion, emotive speech synthesis, and speech emotion recognition