On August 7, Alibaba officially introduced CosyVoice Studio, marking China's inaugural foray into a one-stop AI voice productivity platform. Constructed upon its proprietary large voice model, Qwen-Audio, this platform seamlessly integrates three pivotal functionalities: CosyFlow, a voice recording tool; CosyAgent, a voice agent; and CosyCreative, an audio content creation tool. These features cater to a wide array of scenarios, spanning personal office tasks, enterprise interactions, and audio content production.

CosyFlow stands out by automatically eliminating superfluous spoken words, generating standardized texts such as emails and meeting minutes, and employing voiceprints to differentiate between speakers. Meanwhile, CosyAgent facilitates the creation of real-time conversational agents through natural language, specifically tailored for intelligent customer service and telemarketing scenarios.

Furthermore, CosyCreative boasts a vast library of thousands of built-in voice timbres, supports voice cloning capabilities, and enables users to upload documents to effortlessly generate conversational podcasts or multi-character audiobooks. At present, CosyVoice is accessible via iOS, Android, Mac, and Windows app stores, providing all users with a limited-time free trial.

The two enterprise-grade functions, CosyAgent and CosyCreative, are currently undergoing focused testing through a whitelist invitation mechanism. They are slated for a gradual rollout to all enterprise users in the near future.