K-Content News

AI startup Twelve Labs unveils video language AI foundation model Pegasus-1

  SEOUL, Nov. 9 (Yonhap) -- Twelve Labs, South Korea's generative artificial intelligence (AI) startup, unveiled its video language foundation model Thursday in a bid to take the lead in the hyperscale large language model market.

The model, named Pegasus-1, can accurately summarize long videos into a text and have a chat about the video with its users, offering 61 percent higher performance compared with the existing most advanced video language model, according to Twelve Labs.

The company said Pegasus-1 has been trained on over 1 billion image-text pairs and 35 million video-text pairs, which is about 10 percent of the 300 million diverse video-text pairs the company has collected.

The Korean company said Pegasus-1 can be commercialized immediately, especially in sports, media, entertainment, education and physical security, among other areas.

Twelve Labs, founded in 2021, has attracted a combined US$10 million investment from U.S. chip design company Nvidia Corp., U.S. chipmaker Intel Corp. and two others last month.