Alibaba Launches China's First AI Voice Platform CosyVoice Studio, Integrating Semantic Understanding into Voice Capabilities

On August 7, Alibaba officially launched CosyVoice Studio, China's first one-stop AI voice production platform, built on its self-developed voice model Qwen-Audio, which achieved first place globally in speech recognition (ASR), real-time interaction, and speech synthesis (TTS) on the authoritative AI evaluation platform Artificial Analysis. The platform includes three main features: the voice keyboard converts speech into structured text by removing filler words, the audio content creation tool transforms user content into podcasts and audiobooks, and the voice agent enables businesses to create real-time conversational agents. The platform is the first time Alibaba has aggregated its voice model capabilities into a product, allowing enterprise users to test the experience and then call APIs as needed, while individual users can use it directly. The voice keyboard supports intelligent transcription of special text like numbers and formulas, and its 'notes' mode is designed for long meetings, interviews, and classes, with real-time transcription, speaker diarization, structured chapters, and multi-language translation. The agent tool, CosyAgent, lets users create agents with natural language, supporting deep configuration of prompts and workflows for customer service and telemarketing. The content creation tool, CosyCreative, offers thousands of voices and voice cloning to generate conversational podcasts and multi-role audiobooks. The platform is available for free on iOS, Android, Mac, and Windows via app stores, while CosyAgent and CosyCreative are being tested with enterprise users on a whitelist basis.