[Digital Today reporter Jin-ho Lee] Inserting product images into your finished video allows them to blend naturally into the flow of the video. The brightness and color change depending on the surrounding lighting. Naturally, when people pass by, some of the products are not visible. Without additional filming or complex manual work, AI analyzes the space and movement in the video and synthesizes the images as if they were there from the beginning.
This was made possible by “InsertAnywhere,” an AI video compositing technology jointly developed by SK Telecom and KAIST. Composition tasks that would take days or weeks on a production set can now be completed in just hours.
Na Tae-young (나태영), head of SKT’s enterprise solutions development team, said in a recent interview with Digital Today that AI tools to automatically synthesize advertising images have existed for some time, but in reality, much of the work was often done manually by humans. InsertAnywhere is different in that the output is actually “instant”.
◆Read the flow of the video and synthesize it naturally
InsertAnywhere is a technology that naturally inserts products, props, characters, brand logos, and more that couldn’t be captured during the original video shoot. Users specify where they want to place an image, and the AI applies it throughout the video.
It’s different than just adjusting the image colors and pasting it into a video. Using “4D scene understanding” technology that adds the passage of time to the 3D spatial structure, we analyze camera and object movement, surrounding lighting, reflections, shadows, etc. The core of InsertAnywhere is to naturally blend images and videos created in different environments into a single scene.
For example, if you insert an image of a car into your footage, reflections tailored to the surrounding lighting and space will also be rendered on the vehicle’s glass and surfaces. In a video where a fashion model tries on a dress, it is also possible to change only the outfit and dress the model in a different product. Clothes that apply AI blend naturally with the model’s movements and camera work, just as if the model were actually wearing them.
Product images and company logos used in advertisements must be clearly visible, but if they stand out too much, they can break the immersion. SKT also offers the ability for users to select composite strength and color tone, allowing advertisers to adjust exposure while allowing producers to maintain natural-looking videos.
◆ From weeks to hours…More than 90% of work is automated
InsertAnywhere was developed as the core technology of SKT’s virtual product placement (VPPL) solution AdFlux. Insert Anywhere began joint research in 2025 with the KAIST research team led by Professor Joo Jaegol. A related paper has been accepted to ECCV 2026, a leading computer vision conference to be held in Malmö, Sweden, in September. SKT has also completed patent applications for related technology.
AdFlux, which uses InsertAnywhere, is used in broadcast programs such as “Gukkajiganda Tobak Tour”, “Urichi Geum Manna”, and “Bulmyeong Onjaedeongol”.
Less difficult compositions such as logos and signboards can be completed within a few hours. Although it takes time in scenes where the product is intricately intertwined with other objects or appears for a long time, it can significantly reduce production time compared to manual production.
Mr. Na said that he had heard of cases in the past where people requested AI synthesis but did not receive results even after 15 days. With InsertAnywhere, simple tasks can produce results within an hour or two, reducing work that previously took weeks to at most two days, he said.
He added that SKT believes it has automated more than 90 percent of its current synthesis processes. The rest is for checking whether the AI’s output matches customer preferences and requirements and making corrections, he said.
◆ Expanding beyond advertising to include movies, commerce, and YouTube
SKT plans to expand InsertAnywhere into video production fields other than advertising. When applied to the post-production of movies and dramas, it can add necessary props after filming or insert objects from past eras that are difficult to obtain today. This reduces the need to create or purchase all the props needed to produce historical and period dramas, potentially allowing independent films and small to medium-sized production companies to reduce costs while improving the quality of their videos.
Mr. Na said that video and film production still requires a high proportion of tasks to be handled one by one by humans, which places a burden on small and medium-sized content creators. He said he wants to make the technology available to independent films and small production companies, making it easier for everyone to create content.
Commerce is also a potential market. By sequentially applying multiple clothing images owned by the seller to a video once shot by a model, the cost of casting and re-shooting a model for each product can be reduced.
SKT is also testing a way to use AI to extract product images when you enter the address of an online shopping mall and change the clothes of people in videos. We are considering upgrading this function and providing it to commerce businesses.
Interest is also growing overseas. Na said that after the paper and technology became known, he received a proposal from Tolkier to try a business targeting local broadcasters and influencers. He said SKT is discussing various possibilities for cooperation.
◆API and SaaS are also available… “Lowering the hurdles to compositing images into videos”
Currently, InsertAnywhere is provided within AdFlux, but if we see market demand, we may offer only the core functionality as a separate application programming interface (API). SKT also leaves open the option to expand into a software-as-a-service (SaaS) or editing tool for video creators, where customers upload images and videos, specify only the insertion location, and receive the output.
Na said that SKT is technically ready to provide InsertAnywhere’s core functionality via API. If there is demand, there are no major technical barriers to offering it as a separate service through SK Group’s open API platform, he said.
However, if this technology is to be made available to general users, it is necessary to prepare for abuse such as deepfakes and unauthorized synthesis. SKT plans to prevent such side effects in future open services through login and user authentication, watermarks, and notices of terms of service.
SKT’s video compositing technology development is based on AI capabilities and media business as a commercial foundation. In addition to operating IPTV and broadcasting businesses through SK Broadband, it also has a business foundation that can apply technologies such as advertising, commerce, and video services.
Mr. Na said that although SKT is a telecommunications company, media is also one of its main business areas. He said there are many areas where AI video technology can be applied across the company’s businesses, including SK Broadband, advertising and commerce.
He said that while the technology was initially developed for the specific market of virtual advertising, its applications are endless across video production, film and commerce. The goal is to expand the technology to something that more people can actually use and ease the burden of content creation, he said.
