Sunday, November 15, 2026 Independent reporting on artificial intelligence Newsletter
AI News Beats
Models & Research

Tencent launches Hy Realtime voice model with search and singing

The model is already used in Yuanbao voice calls and Tencent customer service, while a Tencent Cloud API remains a planned release.

Audio equipment with connected cables and green lights in a dark room
Photo: Tyler / Unsplash

Tencent introduced Hy Realtime on November 13, adding web-search decisions and singing to a real-time voice model already used in its Yuanbao app and customer-service products. A Tencent Cloud API is planned, rather than confirmed as generally available.

Chinese business publications Yicai and Jiemian both reported the launch and the distinction between current product use and the forthcoming developer service. That gives the announcement two audiences: people trying a voice assistant now and businesses considering whether to build it into their own software later.

Conversation that can reach beyond the model

IT Home’s report, which reproduces Tencent’s description, says Hy Realtime can use conversational context to decide whether to search online. It also describes handling incomplete speech and corrections during a conversation. These are company descriptions of intended behavior, not findings from an independent test.

For a voice assistant, the difference between speaking and acting matters. Producing a fluent answer does not establish that the answer is current. A search step can bring in outside information, but the assistant still needs to select a relevant source and communicate what it found accurately.

That makes the decision to search an important part of the product, rather than a background detail. A question about a recent event and a request to explain a familiar concept may call for different handling. Tencent’s announcement does not establish how often the system makes that distinction correctly in everyday use.

Where users can try it

IT Home says the current version can be accessed through voice calls in the latest Yuanbao app by selecting the Chuyu voice. It also links to Tencent Cloud’s Hunyuan voice demonstration, which presents a web entry point for trying some capabilities.

The same report describes singing developed with Tencent Music’s Lingyin AI team. Users can request music through conversational preferences. That expands the interface beyond a conventional spoken question and answer, but it should not be confused with a general license to reuse musical output commercially.

Access inside an existing consumer app is not equivalent to an enterprise integration contract. A business considering the model would still need the developer service’s availability, usage limits and commercial terms before planning a production rollout. The launch reports reviewed here do not establish a general API price.

What still needs evaluation

Jiemian’s account places the model in customer service as well as Yuanbao. Those settings involve different expectations: casual conversation can tolerate an interruption or a playful response, while a support workflow needs a reliable answer and a clear way to reach a person when the system cannot help.

No independent comparison in the sources reviewed establishes that Hy Realtime is better than rival voice systems. The next concrete milestone is the promised Tencent Cloud API, followed by documentation that lets developers assess supported scenarios and deployment constraints. Until then, the clearest confirmed change is the model’s introduction in Tencent’s existing voice products.

Stay on top of AI

Get the stories that matter in artificial intelligence, straight to your inbox.

Free. One email with the stories that matter. Unsubscribe any time. See our privacy policy.