DeepBrain AI to Independently Carry Out SME Ministry R&D Project, Develops 'AI Video Agent'

DeepBrain AI has begun work on developing an 'AI video agent.' On the 19th, the company announced that it had formally launched related development after being selected as the sole lead organization for the Ministry of SMEs and Startups' 'SME Technology Innovation Development Project (Global Leading Technology Development)' task.

The project is titled 'Developing an AI Video Agent for Automatic Content Generation Based on a Video Foundation Model.' It will be carried out independently, without any joint or entrusted institutions.

The research period runs from September 2026 to August 2028. The project is worth a total of KRW 1.5 billion, including KRW 1 billion in government-funded R&D support.

The scope of this development covers support for video generation, modification, and editing. The AI video agent aims to go beyond the level of video generation and carry out conversation-based planning, scene composition, and editing. The technologies to be developed consist of a video foundation model, advanced avatar synthesis and lip-sync, and the AI video agent.

The video foundation model analyzes a wide range of materials, including documents, web pages, images, video, and audio, and uses them to generate scenarios, storyboards, and scripts. The video production support areas are marketing, education, commerce, and news. The model plans to reflect corporate logos, brand colors, and product information, and will apply retrieval-augmented generation (RAG).

Avatar technology focuses on advancing DeepBrain AI's accumulated capabilities in its AI human business. The capabilities being advanced are synthesis and lip-sync. Through this, it aims to produce natural mouth movements and facial expressions when speaking multiple languages, while applying movements and expressions that reflect the characteristics of each field, including education, news, and commerce.

The AI video agent connects the video foundation model with avatar technology to automate the entire video production process. In this process, multiple AI systems divide tasks among themselves and handle scriptwriting, scene composition, avatar creation, subtitle insertion, background music insertion, and editing. Users can issue modification requests through text-based conversation.

DeepBrain AI will gradually develop the AI video agent so it can be integrated as a core feature of its video production platform, 'AI Studio.' The company outlined this schedule and implementation plan. This year, it will focus on analyzing requirements in 4 areas, designing core technologies, and conducting initial verification.

It then plans to unveil an integrated beta version linked to AI Studio in 2027, and in 2028 it plans to conduct integrated demonstrations and official testing by field before applying official features to AI Studio. The company will continue development according to this annual rollout plan.

This initiative is tied to the goal of supporting companies that do not have professional editing staff. DeepBrain AI plans to help these companies produce videos for promotion, education, product introductions, and news by using their own materials. DeepBrain AI CEO Jang Se-young said that video production is shifting from a process centered on operating complicated editing tools to one in which users complete outputs through conversation with AI, and that the goal of this project is to build an environment that can efficiently handle the entire process from planning and production to modification and distribution.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408388

References

• https://www.epnc.co.kr/news/articleView.html?idxno=408388