Tencent Hunyuan open-sources the Hy-SimulMT simultaneous interpretation large model, with a response time of about 3 seconds
Tencent Hunyuan has released and open-sourced the simultaneous interpretation large model Hy-SimulMT, which supports real-time Chinese-English speech-to-speech translation and real-time voice cloning, and is now available for free on the Tencent Hy Translation App. According to the official statement, its measured response time is about 3 seconds, compared with an average of 4-5 seconds for professional human interpreters. On the long-audio open-source evaluation set RealSI, the model achieves the best results among open-source models at the lowest latency, and is said to surpass most commercial simultaneous interpretation APIs; in comparisons with 6 CATTI-certified simultaneous interpretation experts, it achieved higher translation quality scores and lower latency.
Why read this
AI score 73
The original text provides the response latency, performance on evaluation sets, and comparisons with human interpreters, allowing readers to judge the usability of real-time speech simultaneous interpretation.
Topics
Body · AI translation
Tencent Hunyuan Simultaneous Interpretation Model Hy-SimulMT Open-Sourced, Response Time Only 3 Seconds, Surpassing Human Interpreters
2026-10-26 22:09 IT Home - Haomiao
Thanks to IT Home users Xichuang, Qingyan, Qingchuan Wenhe, and Huanan Wu Yanzu for the tip-off!
IT Home, October 26 news: Tencent Hunyuan's simultaneous interpretation large model Hy-SimulMT was released and open-sourced today. It features low latency and high accuracy, and supports real-time speech-to-speech translation between Chinese and English — the speaker's audio can be input into the Hy-SimulMT model in real time, and the model translates while listening, providing the translation and speech in real time. It also supports real-time voice cloning, which can highly restore the speaker's voice and tone. The model makes its debut in the Tencent Hy Translation App, offering a free service experience.
According to the official introduction learned by IT Home, low latency is a major highlight of the model. The measured response time of the Hunyuan simultaneous interpretation large model Hy-SimulMT is around 3 seconds, while the average level of professional human interpreters is 4-5 seconds. Evaluation results show that the model's translation performance exceeds mainstream open-source simultaneous interpretation models on the market, and even surpasses most commercial simultaneous interpretation APIs.
Officials say that on RealSI, a commonly used long-audio open-source evaluation set in the industry, Hy-SimulMT demonstrated good simultaneous interpretation performance, as shown in the figure below:
StreamLAAL on the horizontal axis represents the latency metric of simultaneous interpretation, reflecting the number of seconds of delay in the model's output translation results compared with the input audio — the lower, the better. The vertical axis represents translation quality, scored using GPT6.1-Sol — the higher, the better. Hy-SimulMT achieves the best results among open-source models at the lowest latency, and even surpasses most commercial APIs.
On short-audio test sets, Hy-SimulMT also performs well, as shown in the figure below:
To more accurately evaluate the true level of Hy-SimulMT, the team, within the scope permitted by compliance, extensively collected newly emerging Chinese and English audio on the Internet, covering ten fields in total: software AI, medicine, manufacturing and energy, natural sciences, agriculture and food, business management, law, education, culture and tourism, and sports and life, involving multiple scenarios such as speeches, interviews, and meetings.
On this data, the team not only tested the performance of Hy-SimulMT but also invited multiple human simultaneous interpretation experts to perform simultaneous interpretation on these audio recordings.
There were 6 human simultaneous interpretation experts in total, of whom 3 hold CATTI Level 1 Interpreting certificates and 3 hold CATTI Level 2 Interpreting certificates; all experts have years of practical simultaneous interpretation experience. The evaluation compared the translation content of Hy-SimulMT and the human simultaneous interpretation experts in terms of latency and translation quality, with the highest score among the 3 selected for the human simultaneous interpretation experts. In this capability positioning evaluation, Hy-SimulMT reached the level of human simultaneous interpretation experts in translation quality (with a higher score) and was faster (with lower latency).
Advertising Statement: External redirect links contained in the article (including but not limited to hyperlinks, QR codes, passcodes, etc.) are used to convey more information and save selection time; the results are for reference only. All articles on IT Home containing external links include this statement.