MAIU (Media AI Understanding) is an AI video analysis service that helps find desired scenes instantly. By using this service, video editing time can be reduced to one-sixteenth of what it would take when done manually. The expense of the service has also been set at one-third of competitors' rates.
Naver Cloud held a media briefing on the afternoon of the 22nd at COEX in Seoul, introducing its new AI video analysis service, MAIU. This service was first revealed at the International Broadcasting, Media, Audio, and Lighting Exhibition (KOBA 2025), which took place at COEX on the 21st.
MAIU is similar to Microsoft's video indexer, characterized by its use of artificial intelligence (AI) to significantly reduce the video analysis time required during the editing process compared to when it is performed manually. According to Naver Cloud, analyzing 60 hours of raw video by a person would take 32 hours, but using MAIU reduces that to just 2 hours.
Wi Dong-yun, a leader at Naver Cloud responsible for MAIU's development, noted, "In the past, videos were processed one by one at the image level, but now we analyze continuous scenes in segments to eliminate unnecessary duplicate calculations, thus improving analysis efficiency by about 70% compared to the previous program operations." He explained, "This allows for cost reduction, enabling us to set service prices at approximately one-third of competitors' rates." According to Naver Cloud, the expense for the MAIU service is planned to be set at under 10,000 won per hour.
Naver Cloud emphasized that MAIU is a service that enhances convenience and search capabilities. Wi noted, "Through multimodal AI, which analyzes various forms of data simultaneously (text, images, voice, etc.), comprehensive search of video data within the segments has become possible." He added, "For instance, even if one searches for 'find the scene where Kim Jong-min is smiling in the garden and talking about the wedding,' it can be accurately retrieved." He then mentioned that this is possible because the videos have been structured into databases to facilitate intuitive searching and editing by users.
MAIU also offers a 'vision search' function that allows searches based on visual information, as well as an 'audio search' function by person. Wi stated, "MAIU finds matching segments immediately when searching for keywords like 'sandwich' or the action keyword 'dance,'" adding that speaker recognition functions for audio per segment allow for quick searches of speakers' voices and that it particularly provides excellent speech recognition models for the Korean language.
MAIU is expected to be utilized for customized content creation. Seo Ji-won, a manager at Naver Cloud responsible for MAIU planning, said, "MAIU provides a metadata correction function using AI. It supports the entire workflow from video analysis to editing," further explaining that various video edits can be made in the desired direction using a single piece of content, facilitating the provision of customized content.
The MAIU service is currently in closed beta service (CBT) for domestic broadcasters and partners, and is slated for official launch on June 19.