From film promotion to micro-dramas, AI-generated videos boom in China


The promotional video for Venom: The Last Dance is in the style of an animated Chinese ink painting and shows the fictional superhero shapeshifting into various forms. - SCREENGRAB FROM WECHAT

BEIJING: On Oct 23, the day Venom: The Last Dance opened in Chinese cinemas, the official Sony Pictures WeChat account released a 30-second promotional video in the style of an animated Chinese ink painting.

It shows the fictional superhero from the Marvel comic books shapeshifting into various forms, such as a crane and a Chinese lantern, and perching on a pagoda.

What was unusual for a major Hollywood production was that the clip was generated by artificial intelligence (AI).

The technology used is a self-developed software released in April 2024 called Vidu by Beijing-based company, Shengshu Technology. The start-up was set up in March 2023 and is run by former researchers from Tsinghua University, a breeding ground for China’s top AI talent.

In an interview with The Straits Times, Shengshu founder Tang Jiayu – whose firm in June reportedly received hundreds of millions of yuan in pre-Series A funding – said the goal was to showcase efficient video production without sacrificing creative freedom.

Sony bought the copyrights to more than 200 Chinese ink paintings, which were used by Vidu to generate the video.

Tang said: “The resulting video was one that not only demonstrated Vidu’s powerful generation capabilities, but also the myriad new possibilities available to today’s storytellers.”

A Guangdong-based user, commenting on the video, said: “Lively, agile, like a dance – such an ink painting style really suits Venom.”

In recent months, China has experienced a boom in AI video-generation technology. Not only are start-ups like Shengshu attracting investor attention, but tech giants such as ByteDance and Kuaishou are also getting in on the act with their own video generators.

This boom is believed to have been sparked by the release of video generator Sora by US firm OpenAI in February 2024. The generator has not been publicly released for experimentation and testing.

The firm’s launch of ChatGPT in late 2022 brought the potential of generative AI technology into public consciousness.

Similarly, Sora’s launch created debate on the possible uses for AI-generated videos, which could drastically cut time and costs compared to traditional video production, which require camera equipment, film crews, actors as well as elaborate sets.

While AI-generated videos and avatars have been used in online shopping, entertainment could be the next big wave.

In an August 2024 report, Soochow Securities – a Suzhou-based firm providing financial services – estimated that the AI video-generation industry in China could be worth up to 585.8 billion yuan (US$80 billion), covering movies, television dramas, cartoons and micro-dramas.

There was rapid development in AI video-generation models in the first half of 2024, with improvements in areas such as generation time, resolution and frame rate. This means that the gap between Chinese video generators and Sora has narrowed, the report added.

The boom in such content has also caught the attention of the authorities.

In September, the Cyberspace Administration of China issued draft regulation for the labelling of AI-generated content, which requires such videos meant for public consumption to be “explicitly identified”, such as by watermarking or audio prompts.

The regulation is expected to take effect in the coming months. But for now, the government appears to be taking a light approach, especially with its aim of being a world leader in AI by 2030.

Also in September, Shanghai-based MiniMax created buzz when it publicly launched a new text-to-video generation model called Video-01. It was praised for its short clips with realistic human motions, such as running and conversing, actions that earlier video generators struggled with.

The company, backed by tech giants Alibaba and Tencent, is known to investors as one of China’s four “AI Tigers” for their potential to reach the forefront of this fast-iterating technology.

Shanghai-based CreativeFitting is already making headway in commercialising AI-generated micro-dramas overseas, with aims to capitalise on the growing market for this relatively new entertainment format.

On its app, Reel.AI, viewers pay on a subscription basis or for each video, which can be as short as a minute. Such shows are typically created in a vertical format – ideal for viewing on phones – and feature dramatic plot twists to keep viewers hooked.

The app was launched in January 2024, and in July, it was among the top 100 paid entertainment apps on Apple’s App Store. It is currently available in more than 20 countries but its users come mainly from the United States, Britain, Canada, Australia and New Zealand.

Jacky Lin, a partner at CreativeFitting in charge of strategy and financing, said that based on viewer feedback, the fact that audiences knew the dramas were AI-generated did not hamper their enjoyment of the plot and storyline.

“That gave us confidence... (as) we don’t actually aim to compete with other AI video generators, such as in terms of the maximum length that can be produced. We pay more attention to how to tell a good story,” he told ST.

Tech analyst Rui Ma, founder of the Tech Buzz China newsletter, believes AI-generated video content is much more mature in China than in the US. She gave the example of FancyTech, which uses AI to generate advertisement videos for firms on e-commerce platforms.

It benefited from Alibaba asking merchants to use video to market their products on the popular Taobao platform as early as in 2017, while Amazon was slower off the mark in adopting this.

“So not only does FancyTech have more training data, it also has a big market of merchants being trained to use video already,” she said, adding that existing videos help AI models generate new ones.

Even in an increasingly competitive field, the excitement looks set to continue, at least for next year, said Shengsu’s Tang. He believes there is still room for rapid adoption and growth.

“From our perspective, it’s still too early and premature to consider any (industry) consolidation happening, at least within 2025.” - The Straits Times/ANN

Follow us on our official WhatsApp channel for breaking news alerts and key updates!

China , AI , video

   

Next In Aseanplus News

New finance target in COP29 outcome could help developing countries set stronger targets, says Singapore minister
The International Criminal Court is facing pushback and doubts as its member states meet
Man thrown off car bonnet: Former fastfood outlet employee jailed for attempted murder
Laos education ministry reports school dropout rate drops to 11.5 per cent; more students are completing basic studies programmes
Cambodian Information Minister urges action against fake news and AI challenges; warns these issues may hurt South-East Asia in future
Macau’s new leader unveils team to govern gambling hub, with focus on diversifying economy
Shouting match erupts over monsoon tourism proposal in Parliament
Vocational college student's murder: Victim was hit more than once, says friend
Asean gears up for Year of Skills 2025
Indonesia likely to delay start of VAT hike amid growing worries over living costs

Others Also Read