Live Updates > Live Update Details
2025-12-15 16:17:34
[Alipay Tongyi Open-Sources Multiple Speech Models, Launches Lightweight Solution Supporting Local Deployment] ⑴ Alipay Tongyi recently announced the open-sourcing of the Fun-CosyVoice3-0.5B model, which features zero-sample timbre cloning capabilities. ⑵ Users only need to provide a reference audio clip of 3 seconds or more to replicate its timbre and synthesize new speech. This model supports local deployment and secondary development. ⑶ In addition, Tongyi also launched a lightweight version of the Fun-ASR-Nano model, with its total parameters compressed to 800 million. ⑷ This lightweight version has lower inference costs and is now open-source, also supporting local deployment and customized fine-tuning.
