Alibaba Cloud Upgrades Tongyi Bailing Models, Reigniting Competition in Voice AI
Want to read in a language you're more familiar with?
Alibaba Cloud has upgraded its Tongyi Bailing voice AI suite, open-sourcing compact models for on-device use. The update focuses on real-world needs: mixed-language recognition, professional terminology, and cross-lingual voice cloning.
Alibaba unveiled major upgrades to its Tongyi Bailing voice model lineup on December 23, 2025.
The release includes Speech Recognition (Fun-ASR), Text-to-Speech (Fun-CosyVoice3), and End-to-End Voice Interaction (Fun-Audio-Chat) models, with lightweight versions fully open-sourced.
Fun-ASR supports 31-language mixed speech recognition and introduces a RAG mechanism. Fun-CosyVoice3 enables cross-lingual voice cloning. The open-sourced small-parameter models make high-performance speech AI feasible for on-device deployment.
Source：GEEKPARK
