This year’s Asian Media Summit featured a variety of workshops for broadcasters across the Asian region.
Workshop topics include AI, digital newsroom, transmission, sustainability, disinformation and youth.
In one of the day’s workshops developed by Steve Ahern for AIBD and IPPTAR, around 30 broadcasters learned how to apply the latest AI tools to radio, television, podcasts and drama productions.
Demonstrating the power of AI and creating proper prompt instructions, Ahern took group shots and used AI to convert them into realistic video footage. Original authentic group photos are available above. The video below is not authentic. “We’re starting with this still-consuming picture of a group of people and turning it into a 10-second video where people cheer and clapping, and when we clap, the sun and sparks come out of our hands.”
The advantages and disadvantages of using AI include warnings that AI tasks use more energy than regular searches and are not necessarily environmentally friendly.
As responsible use of AI is developed, professional media outlets that plan to use it must keep suppliers accountable for their energy use and use it efficiently so that they are not wasted resources or staff time. Training is required to guide staff to responsible use of AI.
“It’s an exciting new tool, but we need to note that every AI request is very power hungry…
“There are other things to be aware of: publishing a lot of information via AI may not be a good idea for reasons of security and competitive advantage.
Low-cost, open and accessible tools demonstrated in training sessions include:
Research: Editing and cloning ChatGpt, Perplexity, Claude, Deepseek, Grok Audio: CleanVoice, Description, Meta Audio Craft Photo Cloning and Editing: ChatGPT, Creating and Editing Confusing Videos: Description, Flow, Sora
Higher levels of more expensive specialized tools have also been demonstrated, including:
Vida for archives, create keyframes for archives and shots in Midjourney, ImageFx, Deepmind, Photoshop video generation Kling, ruma Ray2, Higgsfield, Veo3 video generation Topaz Lab, Davinci Resolve, Adobe Premier, After Effects for production Enhangment Enhangment Izotope, Cryomix, ai cloned for ai mixing for ai cloned for ai cloned for cryomix for cryomix for roex sound effects
Ahern said one trend for more cost-effective tools is the payment model “as you go” based on the minutes you use, rather than monthly or annual subscriptions.
Translation is also an important AI use case at present, but its success depends on the amount of linguistic data used to train AI models. For minority languages such as Khmer, a language spoken in Cambodia and nearby countries, there is not enough training data to provide quick and accurate translations. It is likely that broadcasters are the source of language training as the AI model expands to other languages, so it is likely that it will be the station’s profitable to develop their own Inhouse AI translation tools and sell the results.
Two KBS Korean experts in AI audio and video Han Ju-Yeol and Yim Young-Jae discussed the findings of using AI in production. study:
Due to the deliberate limitations of technology that does not create graphic deaths or horror scenes, it is difficult to do violent or horror things with AI.
When creating a historical documentary about medieval battles, Han discovered that generative AI video tools cannot create Korean armor with a real look, as they are primarily trained in the appearance of the armor worn by the Western Knights. Chinese AI tool Kling was the most successful in creating Asian style armor.
To avoid infringing copyright, most AI generators are subject to intentional restrictions on copying actual artwork and designs. This prevented the creation of realistic drama scenes where the Korean flag is shown on the flagpole.
The pair pointed out that many broadcasters and content creators use the tool, but they never established professional workflows to ensure that their teams work consistently and most efficiently.
They also recommended that detailed prompts be required to produce the best results. For example, don’t just ask “create a scene where two men are fighting.” Instead, we’ll explain the following details: “We create a 15-second-long battle scene where two men fight. One man wears a dark hoodie and track pants, the other man wears a blue T-shirt and jeans. One has long hair, the other has short hair and a tattoo. One man is carrying a knife and the other man has chains.
According to Korean experts, ChatGpt is not yet on the professional level, but he has found that he is good at creating prompts for use with other professional tools such as Openart and Photoshop.
Some tools are more accurate to respond to prompts than others. Han discovered that Google’s VEO 3 is “one of the most advanced AI models for rapid responsiveness.”
In the case of audio, one of the most important things AI uses is for the separation of vocals and music. This is used by Paul McCart Nelly on Beatles’ track “at the time.” “It can be used to remaster legacy audio,” Yim said. “Multi-track originals are no longer needed. You can remix them from the finished song and separate the elements,” he said.
In the workshop, two Malaysia-based lawyers, Teeruvarasu K Muthusamy and Vimal Sathiaseelan, explained the legal implications of using AI to generate professional content. Key points included:
Can broadcast organizations claim copyright to works generated using AI? Based on the legal definition of artificial intelligence, this test relates to how the “original” is a task and who is the eligible copyright holder.
The core principle in determining whether ownership of work using AI can be claimed is whether there was “sufficient effort” by humans to make the work unique.
These two points are important to consider, as if media companies plunder most of their human staff and create most of their content with AI, they are unable to claim ownership of content creation because they do not put “sufficient human effort” into a creative process.
World law is still developing in relation to AI copyright. According to U.S. law, anything made by AI can “go to the public domain right away and protect copyright,” but China is moving in a different way. “We consider AI as a tool and owns and uses it by stakeholders, so copyright can be held by stakeholders or companies.”
The first comprehensive legal framework for AI is the 2024 AI Act of the European Union. Other jurisdictions have regulations, but there are no laws yet. “The EU acted early to establish principles and frameworks, and its purpose is to provide a high level of legal protection for content creators and other stakeholders,” the lawyer said.
They proposed that media organizations should develop and publish documented policies and procedures to state how AI is used and ensure that there is human surveillance so that they can assert “sufficient effort” as a basis for copyright.
Speaking at the workshop, Cambodia under the Secretary of State of the Ministry of Information, former journalist, radio presenter and television news reporter Ouk Kimseng highlighted the differences in the way content creators approach AI. Some take professional and ethical obligations seriously, as opposed to unregulated amateur content creators who take professional and ethical obligations seriously, and who don’t know about professional obligations.
Participants evaluated the workshop as a success and left plans to return to the organization to new tools to use and how to reinforce AI to the media business.
Related Reports: Accurate and Timely Access to Information is One of the Fundamentals of a Democrat Society: Asia Media Summit 2025
