近日，由北京大学数学科学学院校友戴正与付博铭联合创立的多语言 AI 语音交互平台 ElevTalk 宣布完成 Pre-Seed 轮融资。本轮投资方包括由 Bass Ventures 与 EO Studio 联合运营的美国超早期基金 Patriot Fund、Xiaoxiao Fund，以及一位百度创始团队成员出身的天使投资人。具体融资金额暂未披露。本轮融资将主要用于提升亚洲语言语音交互能力，并进一步拓展美国及亚洲市场。
随着生成式人工智能加速进入办公场景，语音正逐渐成为 AI 时代的重要交互入口。然而，目前主流语音输入产品大多围绕英语母语用户设计，在亚洲语言识别、多语言混合表达及跨语言办公等场景中仍存在优化空间。ElevTalk 聚焦全球多语言专业人士，提供覆盖语音输入、表达优化及跨语言翻译的一体化语音交互体验，支持英语、中文、日语、韩语、越南语等语言的原语言转写、多语言混合表达，并针对亚洲用户常见口音、人名及专业术语进行了优化。
根据公司内部统一测试，ElevTalk 在韩语和中文语音识别任务中的字符错误率（CER）分别达到 1.55% 和 2.25%。公司表示，在相同测试条件下，产品在韩语及中文识别任务中的表现优于部分国际主流语音输入产品。产品上线以来，在未进行市场推广投放的情况下，已获得超过 1000 名早期用户，核心用户主要来自美国硅谷的韩裔、华裔创业者、投资人及科技从业者。数据显示，核心用户平均每周活跃使用 5 至 7 天，最活跃用户单日语音输入超过 4000 个单词，周活跃用户次周留存率已由约 17% 提升至约 50%。
ElevTalk 由首席执行官戴正与首席技术官付博铭共同创立。戴正毕业于北京大学数学学院，后入选清华大学苏世民学者项目（Schwarzman Scholars），曾任职于麦肯锡，并参与大型 AI 语音产品战略工作；付博铭同样毕业于北京大学数学学院，长期从事低延迟、多模态人工智能模型研发，曾参与打造百万用户规模产品。两人分别负责公司产品战略、全球市场拓展及多语言 AI 核心技术研发。
戴正表示，随着 AI 从信息工具逐渐发展为生产力工具，语音正在成为人与人工智能最自然的交互方式之一。“全球越来越多专业人士每天需要在不同语言之间切换工作，我们希望让 AI 能够理解更多人的语言、口音和表达方式，让多语言用户能够更加自然、高效地完成沟通。”
Patriot Fund 投资人、Bass Ventures 董事 Yang Hyung-jun 表示，全球知识工作正在加速向 AI 驱动转型，语音交互也将成为重要的生产力入口。“我们看好多语言 AI 交互的发展方向，以及 ElevTalk 团队在产品迭代和技术研发方面展现出的执行能力。随着全球跨语言办公需求不断增长，我们期待 ElevTalk 持续提升亚洲语言能力，为更多国际专业人士提供服务。”
据悉，获得本轮融资后，ElevTalk 将继续加强中文、日语、韩语、越南语等亚洲语言及相关口音的识别能力，并以美国市场为基础，逐步拓展韩国、日本、中国及东南亚市场。
Peking University Alumni George Dai and Bob Fu's AI Startup ElevTalk Raises Pre-Seed Funding
ElevTalk, a multilingual AI voice platform co-founded by Peking University alumni George Dai and Bob Fu, has raised a pre-seed funding round from Patriot Fund, Xiaoxiao Fund, and an angel investor from Baidu's founding team. Financial terms were not disclosed.
The funding will support further development of AI voice technologies for Asian languages and accelerate the company's expansion across the U.S. and Asia.
As AI becomes an increasingly important productivity tool, voice is emerging as a key interface between humans and AI. ElevTalk focuses on multilingual professionals, offering voice input, writing enhancement, and one-click translation across English, Chinese, Japanese, Korean, and Vietnamese. The platform also supports multilingual code-switching and is optimized for Asian accents, names, and professional terminology.
According to the company's internal benchmark, ElevTalk achieved a Character Error Rate (CER) of 1.55% in Korean and 2.25% in Chinese, outperforming several leading international voice input products under the same testing conditions.
Since launch, ElevTalk has attracted more than 1,000 early users without paid marketing. Its core users include founders, investors, and technology professionals in Silicon Valley. The company's most active users dictate more than 4,000 words per day, while weekly user retention has increased from 17% to approximately 50%.
George Dai, CEO of ElevTalk, graduated from Peking University and was later selected as a Schwarzman Scholar at Tsinghua University. He previously worked at McKinsey & Company and was responsible for voice product strategy at a leading AI voice company. Bob Fu, the company's CTO, also graduated from Peking University and has seven years of experience in multilingual AI model development.
“Voice is becoming one of the most natural ways for people to interact with AI,” said George Dai. “Our goal is to help multilingual professionals communicate naturally across languages, accents, and cultures.”
Yang Hyung-jun, Director of Bass Ventures and investor through Patriot Fund, said the team has demonstrated strong execution in multilingual AI and global product development.
Following the financing, ElevTalk plans to further improve speech recognition for Asian languages and expand across the United States, South Korea, Japan, China, and Southeast Asia.
（免责声明：本文为本网站出于广告目的进行转载发布，不代表本网站的观点及立场。本文所涉文、图、音视频等资料之一切权力和法律责任归材料提供方所有和承担。本网站对此咨询文字、图片等所有信息的真实性不作任何保证或承诺，亦不构成任何购买、投资等建议，据此操作者风险自担。）
