[Tencent's Hunyuan Launches Hy Realtime Real-Time Voice Model: Speaks, Sings, and Understands Temporary Changes] Beating AI News: Tencent's Hunyuan has released a new real-time voice model, Hy Realtime, adding online search capabilities for singing and voice dialogue. It has already been integrated with Yuanbao Voice Call and Tencent Customer Service. Users can directly use voice commands to have the AI search for the latest information, and can also request songs by artist, genre, language, and release year. For example, asking it to sing an English folk song from 2010 to 2020 will allow the model to select and sing the song based on the context. The singing function was jointly developed by Hunyuan and Tencent Music's TME Lingyin AI team. The new model also addresses the issue of voice AI easily "losing its intelligence when it starts speaking." The knowledge and reasoning abilities originally possessed by large text models may be lost when transferred to voice dialogue models. Tencent redesigned its training data production process, claiming that the effective transfer retention rate of high-quality text data to voice tasks has increased from 21% to 69%, approximately 3.3 times higher than before. Hy Realtime also enhances its understanding of authentic spoken language. When users speak unclearly, only say half a sentence, or change their mind mid-sentence, the model can reconstruct their intent based on the context. When generating responses, it can simultaneously organize text and synthesize speech, reducing waiting time. Currently, users can select the "First Language" voice experience in the latest version of the Yuanbao App's voice calls. The Tencent Cloud API is planned to be opened soon, and full-duplex interaction and long-term personalized memory remain areas for future upgrades. [Original Link]