[Selected Companies for the Catholic Kwandong University Early-Stage Startup Package 2025] Kwon Mingi, CEO of CineLingo

‘CineLingo,’ an AI-powered knowledge content translation startup

Providing localized videos to viewers through a comprehensive treatment of voice and screen elements in videos
Specializing in knowledge-centric videos across fields like education, science, and technology

CineLingo is an AI-powered generative content translation startup. CEO Kwon Mingi (29), who is currently pursuing his Ph.D. in artificial intelligence at Yonsei University, founded the company in June 2023.

Kwon’s research focuses on generative AI, particularly video and image generation and editing technologies. During an internship at Adobe Research, he researched advanced video editing technologies and is currently developing natural lip-sync and character editing techniques. Kwon has translated and localized math and science content while operating Korean-language channels for Veritasium and 3Blue1Brown.

Kwon explained, “CineLingo provides more naturally localized videos by converting the original video’s audio into Korean, applying matching lip-sync, removing existing subtitles, and overlaying new ones. We specialize in knowledge-centric videos, particularly in education, science, and technology, and we aim to help more people access excellent content without a language barrier.”

CineLingo’s core service is AI-based video translation and localization. The company goes beyond simple subtitle translation, comprehensively treating both the audio and visual elements of the video to provide localized content to viewers.

“We accurately convert the speaker’s voice in the video to text and translate the extracted script using a large language model (LLM) to produce natural sentences that fit the context. We generate the translated script using natural-sounding synthetic speech, then edit the audio to match the speaker’s lip movements seamlessly. We remove any existing foreign-language subtitles or graphics from the original screen, replacing them with new elements if necessary. We also recognize and translate text embedded within the video, such as slides and signs, and arrange them again. Currently, this technology is being applied to YouTube channels such as Veritasium (Korean) and 3Blue1Brown (Korean) to localize knowledge and educational content into the Korean language. We plan to expand into other channels and content areas in the future.”

Kwon stated, “CineLingo’s competitive advantage lies in developing core technologies in-house.” “We conducted most of the R&D independently—STT, TTS, lip-syncing, video inpainting, OCR, etc.—excluding the large language model (LLM). This approach allows us to improve quickly or optimize necessary features for specialized requirements.”

CineLingo has profound insight into the content creator and YouTuber ecosystem. “Having operated our science and math channels with hundreds of thousands of subscribers, we understand the quality of translation and localization and the workflows that creators actually need. We provide solutions with real value based on actual operational experience, not just as a technology supplier.”

CineLingo is also strong in AI technology convergence and automation. “Video translation cannot be solved by a single technology. It requires integrating multiple AI technologies spanning audio, text, and video into a single workflow. We design and operate systems that organically connect these elements. One of our additional strengths is our accumulated data and expertise specifically tailored for knowledge-based content. With our background in translating scientific and educational content rather than general entertainment or dramas, we excel at precise glossary selection and contextual handling. These elements combine to create a competitive advantage that goes beyond simple translation, delivering video content in the most natural way to local viewers.”

CineLingo has pioneered its market by directly contacting content creators. “Thanks to our experience as YouTubers, we have a relevant network. In practice, simply sending proposals via email often generates significant interest and frequently leads to collaborations. Recently, we’ve been preparing to offer individual technologies used in our video translation pipeline as standalone services. For example, we’re currently testing a beta version of our internally developed text-to-speech (TTS) technology. We plan to make these services available so that more creators and companies can utilize CineLingo’s technology. We are in the early stages, but we see providing technology services to external parties as a key marketing strategy moving forward.”

CineLingo is operated by four core members. “My fellow Ph.D. student, Jeong Jaeseok, is spearheading the technology and business as a co-founder. A team member who completed his master’s program in the same lab is developing the image treatment pipeline. Additionally, a team member with a bachelor’s degree in computer science has joined to handle the actual service implementation and operations. Furthermore, our lab’s advisor participates as a non-executive director, providing significant assistance with technical consulting and strategic direction.”

Regarding future plans, Kwon said, “In the short term, our goal is to officially launch all the technologies developed thus far as services by the end of this year.” He added, “We plan to create platforms for pipelines such as STT, TTS, lip-syncing, and video inpainting, which are currently only used internally, so that external customers can utilize them.”

Founded: June 2023
Main business: AI-based video translation and localization services; operation and localization of YouTube channels in the science and education fields
Achievements: Successfully operated a Korean-language YouTube channel through official collaborations with renowned international channels, such as Veritasium, 3Blue1Brown, and Welch Labs; Started operating a Japanese-language YouTube channel in July 2025

reporter jinho lee
[email protected]