Motion Power (co-representative Hongsik Cho, Jihwa Lee) announced on the twenty sixth that it had implemented an update that improved the performance of ‘Daglo’, a synthetic intelligence (AI) service for record management.

Daglo is a service that converts and organizes videos of meetings, calls, and lectures into text for greater than 300,000 members. It’s explained that this model upgrade has raised the performance to the very best level within the industry.

There are three major points to enhance performance. Initially, by enlarging the scale of the STT (speech-to-text) model itself, it is alleged that the speech recognition accuracy has improved to shut to 100%.

As well as, by increasing the voice detection (VAD) function, it could possibly accurately separate voices even in conversations between multiple people. Finally, by clearly displaying numbers and English pronunciation, the text conversion leads to improved readability.

In consequence, Daglo explained that each Korean and English speech conversions are at the very best level, and at the identical time, sophisticated speaker separation is supported.

Motion Power said that the largest difference from other voice conversion services is ‘multimodal’ support.

In other words, should you upload a video file or paste a YouTube link, the audio of the video is created as a manuscript. Also, by installing Daglo as a Chrome extension program, you may start converting text with DaGlo while watching YouTube videos. It is alleged that this feature is preferred by users who need to quickly grasp only the vital parts of an extended video or who want to observe a lecture video as a manuscript.

As well as, the conversion result could be downloaded as a subtitle file in addition to a general document file. The function scheduled for the last update in the primary half of the yr is to convert and display call contents directly into text.

Lee Ji-hwa, co-CEO and CTO of Motion Power, said, “AI is helpful in that it saves human effort and time. It is just not a recent and interesting trend, but we’re focusing only on technologies which might be really helpful to people,” he said. did.

He also said, “The goal of Daglo is that conversations, videos, and audio records are routinely organized without human intervention,” and “We’ll develop Daglo right into a real ‘AI assistant’ for on a regular basis life.”

As well as, it was revealed that it’s specializing in commercializing its own large language model (LLM), which is in the ultimate stage of development. Through this, Daglo, which creates the present manuscript, selects keywords, and summarizes, is upgraded, and in the long run, it’s planned to arrange and inform what to do after the meeting and the subsequent schedule, and add a function that selects and summarizes only probably the most watched parts of the video.

Reporter Lim Dae-jun ydj@aitimes.com