Announcement • 4h
Unisound AI Technology Co., Ltd. Completes Comprehensive Upgrade Of Speech Large Language Models U2-ASR And U2-TTS Unisound AI Technology Co., Ltd. has completed a comprehensive capability upgrade for U2-ASR and U2-TTS, continuously strengthening the capabilities of its multimodal large language models and achieving the expansion from localized coverage of more than 100 dialects to multiple international languages worldwide in the domain of speech large language models, which has further solidified its capabilities as a global speech interaction infrastructure and provided efficient, low-barrier technical support for scenarios such as overseas expansion, embodied intelligence, and agent-to-agent interactions. In this upgrade, U2-ASR has supported recognition for 13 new international languages, covering key overseas markets such as Europe, Southeast Asia, the Middle East, and Latin America; U2-TTS supports multilingual voice synthesis for 8 new languages in Southeast Asia. To date, the U2 speech large language model supports over 100 Chinese dialects and more than 15 international languages. Enterprises can process audio content in different languages by connecting to a single large language model and one set of interfaces, significantly lowering the barriers for developing, deploying, and maintaining multi-language speech services. Under unified evaluation standards, U2-ASR demonstrates outstanding performance compared to industry benchmark models: the average Character Error Rate (CER) for 113 languages is only 6.58%. In real-world business scenarios without language tags being input, it maintains high accuracy through its automatic language identification and closed-set routing capabilities, effectively avoiding language misidentification. Meanwhile, in the ChinaVoices Challenge 2026 Chinese Multi-dialect Speech Recognition Challenge, U2-ASR ranked first in the limited data track with a CER of 9.235% and took the second place in the open data track, further validating the leading position of its speech recognition technology. In subjective evaluations of intelligibility and naturalness, U2-TTS outperforms mainstream industry models. It utilizes a streaming neural network acoustic model to achieve chunk-by-chunk real-time output of 24kHz high-fidelity speech, significantly reducing first-packet latency and meeting the demands of highly time-sensitive interactions such as real-time voice conversations. This upgrade enables U2-ASR and U2-TTS to work together seamlessly, offering a complete multi-language speech interaction chain for AI agent use cases for enterprises. Currently, the upgraded U2-ASR and U2-TTS have been fully launched on the TokenHub of the Company, which is a large language model service platform, with standard APIs made available. Announcement • Jul 22
Unisound AI Technology Co., Ltd. Provides Earnings Guidance for the First Half Ended June 30, 2026 Unisound AI Technology Co., Ltd. provided earnings guidance for the First Half ended June 30, 2026. For the period, the company expects to record a total operating revenue of RMB 530 million to RMB 580 million, representing an increase of 31% to 43% compared to the same period in 2025. The revenue growth rate may double compared to the same period in 2025, demonstrating the Company's strong growth momentum. By business segment, for the First Half of 2026, the intelligent agent business powered by large model capabilities expects to record a revenue of RMB 515 million to RMB 550 million, representing an increase of 200% to 220% compared to the same period in 2025. The direct Token invocation business of large model developed newly by the Company this year expects to record a revenue of approximately RMB 30 million for the First Half of 2026, with a gross profit margin of not less than 60%, of which Token invocation revenue in the second quarter of 2026 is approximately RMB 25 million, representing a quarter-on-quarter increase of approximately 500% compared to the first quarter of 2026. The Company's revenue demonstrates strong sustainability. In the First Half of 2026, revenue from repeat purchases is expected to account for over 60% of total revenue, with repeat purchase revenue growing by 40% year-on-year, fully demonstrating the high customer stickiness and revenue continuity of its core business. Recent Insider Transactions • Jul 21
Co-Founder recently bought HK$282k worth of stock On the 13th of July, Wei Huang bought around 4k shares on-market at roughly HK$70.47 per share. This transaction amounted to 15% of their direct individual holding at the time of the trade. In the last 3 months, they made an even bigger purchase worth HK$808k. Wei has been a buyer over the last 12 months, purchasing a net total of HK$2.0m worth in shares. 
