Back home

Hy Realtime

x+web 0.93
Visit source

Intro

Hy Realtime is the new real-time speech model released by Tencent Hunyuan on 13 November 2026. Yicai published at 16:38:55 Beijing time, and National Business Daily at 16:46:38 wrote that the post came from Hunyuan’s official WeChat account. Both restated the official description: the model performs well on speech understanding, instruction following, and more human-like expression, and adds agentic tool use and singing. Hy Realtime is already live in Yuanbao voice calls and Tencent customer service, and Hunyuan said a Tencent Cloud API would follow soon.

IT Home (16:41:34) carried a longer official explainer. According to that account, the model can use conversation context to recover intent from mumbled, unfinished, or mid-sentence corrected input, and it keeps following multi-turn instructions in longer dialogues. Besides ordinary conversation, it can decide from context whether to run a web search, tying together the search decision, tool calls, and the spoken answer. Officials also said spoken delivery is better than the previous generation on audio quality, pronunciation, prosody, and emotion. The singing capability was co-designed with Tencent Music’s TME Lingyin AI team; users can request songs by artist, album, genre, language, release year, and popularity, or continue from earlier turns — for example, “play an English folk song from between 2010 and 2020.” The current version can be tried in the latest Yuanbao app by selecting the “初语” (Chuyu) voice in a voice call, or by opening the web demo at rtcube.cloud.tencent.com. Hunyuan said the model is still being iterated, with next steps including duplex interaction, longer-term personalized memory, and foreground/background agent interaction.

This page did not find a publicly reachable URL for the original WeChat post, so it relies on Yicai and National Business Daily’s restatement of that account, plus the official explainer quoted by IT Home. Do not describe the forthcoming Tencent Cloud API as already generally available, and do not treat Hy Realtime as the same product as the earlier-stored Hy Memory 2.0 / MemHub or Hy4.

Key points:

Explain

Why it matters

This is a formal model name confirmed on the afternoon of 13 November 2026 by Tencent Hunyuan’s official WeChat account. Yicai’s 16:38:55 Beijing-time stamp falls inside this scan window (16:32–20:32), and the previous 16:32 scan did not record it. It is described as a new real-time speech model already wired into Yuanbao voice calls and Tencent customer service, not as an unnamed feature update.

This page only restates Yicai, National Business Daily, and IT Home’s accounts of the official explainer. Claims about intent recovery, web search, singing, and the “初语” voice come from those public reports. This page does not independently evaluate the model.

Limits

The public record is mainly media restatements of an official WeChat post, plus a web demo. It does not give parameter counts, training data, a standalone model card, or a formal Tencent Cloud API document. Do not describe this entry as an API that is already generally available, or as a feature already turned on for every Yuanbao user and every voice. Do not describe Hy Realtime, Hy Memory 2.0 / MemHub, and Hy4 as the same product.

Usage

  1. Read Yicai at https://www.yicai.com/news/103440081.html , and cross-check National Business Daily at https://www.nbd.com.cn/articles/2026-11-13/4647109.html , IT Home at https://www.ithome.com/1/026/825.htm , and Caijing at https://tech.caijing.com.cn/20261113/5193624.shtml ;
  2. To try it, follow the public reports: open the latest Yuanbao app and choose the “初语” (Chuyu) voice in a voice call, or visit the web demo at https://rtcube.cloud.tencent.com/component/hunyuan/index.html for a subset of capabilities;
  3. When citing intent recovery, in-call web search, singing, or the TME Lingyin AI collaboration, attribute them to the 13 November 2026 official Hunyuan explainer as restated by those outlets, and recheck the upstream source;
  4. Do not describe the forthcoming Tencent Cloud API as already live, and do not describe Hy Realtime as Hy Memory 2.0 / MemHub or Hy4; this page only restates public reports — recheck the upstream sources before citing them externally.

Advertisement