AI video content search technology platform Twelve Labs raises investment from global investors

Video content has become an integral part of the new world. With numerous platforms offering video content like YouTube, Vimeo, Tiktok, etc., it isn’t easy to search the content that suits users’ tastes. Korean startup Twelve Labs has built an AI cloud service that helps in video search and understanding. The company’s platform is getting investor attention worldwide.

Twelve Labs has attracted an investment of $12 million (approximately 16 billion won). The company has completed the seed round by attracting a total of $17 million (about 22 billion won) investment.This investment was led by Radical Ventures, an AI-specialized venture capital (VC) and an existing investor. In addition, Index Ventures, which led the last seed investment, Wonder Co Ventures of Jeffrey Katzenberg, former chairman of Walt Disney and founder of DreamWorks, and Spring Ventures, led by Song Wook-hwan, former CEO of Nike Korea, participated as new institutional investors. New angel investors include Jay Simmons, former CEO of Atlassian, Nicholas Dessane, former CEO of Algolia, and Lucas Bayworld, CEO of Waits & Byasis.

Twelve Labs intends to commercialize a super-large AI model with billions of parameters specialized in image understanding by using new funds. It will be a first for a Korean startup to significantly challenge the development of a super-giant AI model, which was thought possible only for big tech companies.

The company has begun aggressively recruiting AI R&D personnel in Seoul and San Francisco offices and recently entered into a large-scale cloud partnership with Oracle to secure computing resources for AI learning. Twelve Labs plans to train this model using thousands of the latest AI training Nvidia A100, and H100 graphics processing units (GPUs) secured through a multi-year partnership with Oracle’s OCI (Oracle Cloud Infrastructure).

The model is expected to automate video-related tasks, such as video classification, summary, and recommendation, in the same natural way humans do, beyond conventional video search. The company believes that the model being developed this time will fundamentally change how humans communicate with video data.

Jaeseong Lee, CEO of Twelve Labs, said, “Video captures our world better than any other form of data. Therefore, building a super-giant AI that understands this at the human level is a critical task, even if it is difficult, that profoundly impacts our lives. Thank you to our customers and Oracle for being with us this time. As a Korean company, we will lead this trend of innovation on the global stage, contribute to Korea’s AI sovereignty, and ultimately contribute to the development of AI beneficial to humankind.”

Twelve Labs, established in early 2021, won the video search category of the ‘Value Challenge 2021’, a world AI image recognition competition hosted by Microsoft and the International Computer Vision Society (ICCV) in October of last year, beating Kakao and Tencent.

Radical Ventures partner Rob Tewth, who led the investment, said, “Compared to the performance in October, Twelve Labs has realized more than twice the search accuracy and performance over the past year. At the same time, by providing this technology to various customers in a commercialized form, it proved its commercial potential. It showed that it is leading the global video AI industry. I look forward to seeing how it will change the industry landscape.”

From April this year, video understanding AI-based video search API solutions for developers are provided in closed beta form. They are applied to the services of companies in the United States. Starting in July, a visualized web platform called ‘API Playground’ was also provided, allowing users to upload desired videos and try searching directly, so even non-developers can easily use video search technology.

Also Read,