CosyVoice Alibaba (2024年)(コージーボイス)
2024年7月Alibaba FunAudioLLM発表CosyVoice・Industry-leading multilingual TTS + Industry-leading 5-language Chinese+English+Japanese+Korean+Cantonese + Industry-leading zero-shot voice cloning + Industry-leading Alibaba FunAudioLLM CosyVoice Apache 2.0 multilingual TTS 2024。
概要
CosyVoice は2024年7月Alibaba FunAudioLLM発表・Industry-leading multilingual TTS 2024年Alibaba FunAudioLLM CosyVoice Apache 2.0 multilingual TTS position確立。Industry-leading 5-language Chinese + English + Japanese + Korean + Cantonese + zero-shot voice cloning + Speech LLM architecture + 170K+ hours training + Apache 2.0 + Alibaba China。
主な特徴・仕組み
- Industry-leading Alibaba FunAudioLLM team (Alibaba DAMO Academy) 2024年7月 release
- Industry-leading CosyVoice multilingual TTS signature
- Industry-leading 5-language signature: Chinese + English + Japanese + Korean + Cantonese
- Industry-leading zero-shot voice cloning signature
- Industry-leading Speech LLM architecture signature
- Industry-leading 170K+ hours training data
- Industry-leading instruction-following voice control signature (emotion + speaker + style)
- Industry-leading cross-lingual voice cloning signature
- Industry-leading streaming inference signature
- Industry-leading CosyVoice 2 follow-up 2024年12月 (post-release evolution)
- Industry-leading Apache 2.0 license signature
- Industry-leading HuggingFace + ModelScope native
- Industry-leading Alibaba 1999 Hangzhou Jack Ma 25 years
まとめ
CosyVoice = 2024年7月Alibaba FunAudioLLM発表のpremium Apache 2.0 multilingual TTS。Industry-leading multilingual TTS + 5-language Chinese + English + Japanese + Korean + Cantonese + zero-shot voice cloning + Speech LLM architecture + 170K+ hours training data + instruction-following voice control emotion + speaker + style + cross-lingual voice cloning + streaming inference + CosyVoice 2 follow-up 2024年12月 evolution + Apache 2.0 + HuggingFace + ModelScope native + Alibaba FunAudioLLM team DAMO Academy + Alibaba 1999 Hangzhou Jack Ma 25 years position確立。