DingTalk AI Voice Input: Alibaba's Qwen-Audio Tops Global Rankings

DingTalk has launched a new intelligent voice input feature, making it the first collaboration application to support full-scenario AI voice input. This comes as Alibaba's Qwen-Audio model family ranks first globally across three categories on Artificial Analysis. The Qwen-Audio model, which underpins DingTalk's new feature, achieved a character error rate of just 1.7%, surpassing other top international models. It secured the number one global ranking in Automatic Speech Recognition, Real-Time Interaction, and Text-to-Speech. DingTalk's voice input now covers all core scenarios on the platform, including instant messaging, documents, and AI spreadsheets. It also supports use in external browsers and desktop AI agent platforms. On the same day, Alibaba launched CosyVoice Studio, an all-in-one AI voice productivity platform. CosyVoice Studio offers modules like CosyFlow, which can filter filler words and generate formatted emails or meeting minutes from dictation. The "Quick Notes" feature in CosyFlow can distinguish speakers and create structured summaries. This means users can now leverage advanced AI voice capabilities across many applications.