Media companies are trying and testing various ways to use AI. At this stage, AI transformation in media can be summed up as active use of AI to cut production costs, open up new genres and strengthen user experience (UX).

AI video tools that initially only generated clips of a few seconds can now produce 4K resolution video with native audio and long-form content of more than 2 minutes. Advances include character consistency that keeps the appearance of figures even as scenes change, and camera controls that precisely specify camera movements.

As major tech companies compete, the media industry's AI transformation is evolving quickly in both quantity and quality.

Google DeepMind's video-generation model Veo has reached a level where it generates dialogue, sound effects and background music at the same time as video. Adobe, known for image editing tools, also released Firefly Video early last year. It emphasises that it can be used for commercial video without IP disputes because it is trained only on licensed content. New York-based AI startup Runway's Gen 4.5 supports multi-shot and audio generation and is targeting studio workflows in the film and advertising industries.

Chinese companies are targeting the market with price and volume. Kling, from major Chinese short-form video platform company Kuaishou, handles 18 tasks, including video generation, editing and conversion, on a single platform. AI startup MiniMax, which listed on the Hong Kong stock exchange in January, offers its video generation service Hailuo for $15 a month, winning support from solo creators and small and mid-sized production companies. It is rated as having high character-focused cinematic video quality for the price.

Production attempts using these tools are also continuing.

South Korea's terrestrial broadcasters saw the potential in election coverage. KBS introduced generative AI into its 2025 presidential election vote-count broadcast, automatically generating in real time candidate graphics and countdown videos based on changes in vote share. MBC restored footage of an independence activist with AI and used it in its opening, and SBS featured an AI character called "Tupyorro". As the use is for large events rather than everyday production, it remains at an experimental stage.

Among private production companies, CJ ENM is active. It unveiled last year "Cat Biggie", a 100 percent AI animation, using its in-house AI video production system "Cinematic AI". The system integrates images, video, sound and voice in a single system and automatically processes 3D characters and environments. CJ ENM plans to expand into producing AI dramas and AI films this year.

China is moving aggressively to commercialise. The 68-episode AI live-action drama "The Strategies of the Empress Dowager and the Little Princess", released on Douyin, a Chinese domestic short-form platform operated by ByteDance, logged a cumulative 210 million views. The sci-fi micro drama "The Sun That Fell", released in April 2025, used AI to handle the entire process of a 30-episode series with more than 50 characters and 200 scenes, and the production period was 3 months. The structure allows a 10-person team to complete 100 minutes of content within 10 days. Production costs are about one-fifth of the traditional method.

South Korean production companies have also started experimenting. Short-drama platform Vigloo released "Bloodbound Luna" in March, a short drama produced entirely with AI. It is a 22-episode dark fantasy series with fewer than 10 production staff involved. EO Contents Group is producing a seasonal AI drama titled "Soon, It Will Be Night", consisting of 10 episodes of 5 minutes each.

By contrast, the United States, the world's largest entertainment market, has strong resistance from creators. SAG-AFTRA, a union for U.S. actors and voice actors, held a large strike in 2023 with restrictions on the use of AI replicas as a key issue, and later made consent and compensation mandatory in agreements when AI is used. AI video generation technology is advancing rapidly, but it is not leading to production of commercial content.

On YouTube, long-form AI dramas are appearing one after another. In South Korea as well, content that realises fantasy genres that are difficult to film in live action with 100 percent AI is popular. An 18-minute AI drama released by YouTube channel Winkadia has surpassed 470,000 cumulative views. Viewers have responded, "It is better than sloppy live-action dramas."

FROM SCHEDULING TO RECOMMENDATIONS: AI DUBBING BECOMES ESSENTIAL FOR GLOBAL DISTRIBUTION

In content distribution, AI has also become core infrastructure. Personalised AI recommendation algorithms are replacing scheduling.

According to Netflix, 75 to 80 percent of total viewing time comes through algorithm recommendations. Only 20 to 25 percent comes from viewers searching for content. OTT users are known to typically skim 10 to 20 titles within 60 to 90 seconds, and if the content they want to watch is not surfaced within that time, the likelihood of churn rises.

AI algorithms are also applied to thumbnails. Even for the same content, different images appear for different users. Netflix assigns 70,000 to 80,000 tags to a single piece of content, covering factors such as genre, emotional tone, type of ending and protagonist traits. It combines tags with viewing history to select and deliver the thumbnail most likely to be clicked by each user. The same drama is shown to romance fans with couple scenes and to action fans with chase scenes.

From the perspective of global distribution, AI dubbing is emerging as new competitiveness. AI dubbing tools that handle voice replication, preserving emotional tone and synchronising lip movements (lip-sync) are appearing and lowering language barriers for content.

Social media adopted it quickly. After YouTube announced an AI automatic dubbing feature in late 2024, it fully opened it in February this year as a service for all creators supporting 27 languages. Meta officially launched automatic dubbing and lip-sync features last August on Facebook and Instagram Reels. The trigger was YouTuber MrBeast, the most-subscribed in the world, pointing out directly to Mark Zuckerberg the lack of multilingual support.

Netflix is also incorporating AI into its dubbing process. Typically, Netflix originals are translated into languages for 30 countries, and general content into languages for 10 countries. AI generates a first draft of a translation and professional translators refine it. AI is also used to optimise lip-sync to match mouth movements with dialogue. The voices themselves are still performed by people. According to a Netflix official, as many as 1,320 people were mobilised to dub Squid Game Season 3.

While global platforms internalise AI dubbing, South Korea is entering the market through startups and government support programmes. Eastsoft runs "PERSO.ai", Hudson AI runs "Timber", and VBridge operates its own AI dubbing service.

The government is also supporting the AI dubbing industry. The Ministry of Science and ICT and the Korea Association for ICT Promotion (KAIT) promoted a support project called "AI dubbing-specialised K-FAST expansion". The budget is 8 billion won. The goal is to secure a total of about 4,400 AI-dubbed content pieces, or about 1,400 hours, build 20 K-channels, and then transmit them to 22 countries.

NEED FOR INSTITUTIONAL IMPROVEMENT WITH BALANCED REGULATION AND PROMOTION

As AI changes production and distribution at the same time, concerns about side effects are also growing. There are calls for institutional improvements, especially on copyright ownership, creator protection and review systems.

Copyright for AI-generated works is also an issue. In the 2023 Thaler case, the U.S. District Court for the District of Columbia ruled that AI-generated works without human creativity cannot be granted copyright. The U.S. Copyright Office (USCO) also confirmed its position that copyright law protects only works created by humans. It added a proviso that partial protection is possible if a human substantially contributes and edits. In South Korea, discussions to revise the copyright law are under way led by the Ministry of Culture, Sports and Tourism, but it is not easy to set standards because there is no global standard.

Concerns are also being raised about protecting creators as content production workers such as voice actors, translators and scenario writers are replaced by AI. In South Korea, groups such as the Korea Voice Actors Association and the Korea Scenario Writers Association have raised the issue, but it has not led to protective measures at the level of agreements.

There are also no separate standards to review content when AI-generated works are distributed via broadcasting and online. The AI Basic Act, which took effect in January, imposed an obligation to label AI-generated works, but it does not judge what the video contains. That is why there are calls for the labelling obligation under the AI Basic Act to be linked to the review system.

A media industry official said, "In the field, it feels like the technology we can use and the standards for what we are allowed to use are moving separately." The official added, "A single window that ties together creator protection, labelling and distribution responsibility is urgent, but the agencies that enforce and promote the law are dispersed." Institutions need to be improved at the pace at which AI is reshaping the media industry.