According to ThinkTank Beating monitoring, Alibaba has launched the all-in-one AI voice platform CosyVoice Studio, which is powered by the Qwen-Audio series models, covering voice recording, real-time voice Agent, and audio creation.
CosyFlow focuses on voice recording. As users speak, it automatically removes filler words and repeated expressions such as "um," "uh," and "you know." It can also directly organize scattered dictation into lists, emails, and meeting summaries. During recording, it can differentiate between different speakers, automatically generate summaries and to-dos.
CosyAgent targets enterprise voice services. Users can create real-time conversation Agents using natural language, and they can also integrate enterprise documents, databases, MCP, and APIs for customer service, telemarketing, and other scenarios.
CosyCreative is responsible for audio production. By uploading documents, web pages, or entering text, users can generate podcasts and multi-role audiobooks. It also supports voice cloning, partial regeneration, and multi-track editing.
The personal end is now available on iOS, Android, Mac, and Windows for a limited free trial. CosyAgent and CosyCreative are currently being tested by enterprises in whitelist mode.
