Asia Tech Wire (Feb 19) -- Chinese tech companies have been claiming they have something to do with the video generation feature after OpenAI released Sora, its first text-to-video artificial intelligence model.
Focus Technology Co Ltd (002315.SZ) said it has upgraded again AI Mic, an AI foreign trade assistant for foreign trade companies on its Made-in-China.com website, introducing the video generation feature for the first time in the industry.
It can automatically generate HD videos of up to 45 seconds based on uploaded product pictures, solving the problem of lack of professional video producers, time-consuming and labour-intensive video production for foreign trade enterprises, and saving a lot of operating costs for foreign trade enterprises.
Guangdong Insight Brand Marketing Group Co Ltd (300781.SZ) said the company's AIGC project team plans to develop the text-to-video function in March, and wait for the time to be ripe to put it into beta testing.
As for whether Hangzhou Hikvision Digital Technology Co Ltd (002415.SZ) has products similar to Sora, a source said Monday that the company's products are not in the same category of AI as Sora, and that the company focuses on perceptual intelligence.
The person said Hikvision's large video model is mainly for use in the smart IoT industry, coming to enterprises to reduce costs and increase efficiency.
A source from Zhejiang Dahua Technology Co Ltd (002236.SZ) also said on Monday that the company has the capability of large video models, and is currently doing research and development in two industries.
What's different is that Sora is generative, while Dahua is analytical, the two developing in completely opposite directions, the person pointed out.
Specifically, Dahua mainly takes existing clips or customer content for parsing, and then tells its clients what's happening in the videos, and makes behavioural judgements based on that happening.
A representative of Hanwang Technology Co Ltd (002362.SZ) said on Monday that the company has been ploughing its way through the artificial intelligence sector and has its own core technologies in the field of multimodal recognition technology, such as text recognition, image recognition and video analysis.
The representative said that in the direction of recognition and parsing technology, Hanwang has its own large model, which is mainly applicable in vertical fields.
Staff from Easy Click Worldwide Network Technology Co Ltd (301171.SZ) revealed on Monday that the company is already using AIGC technology to provide some creative material generation services for its customers, hoping to benefit from the new model, Sora, to achieve cost reduction and efficiency.
"The company's AIGC creation platform, KreadoAI, mainly generates digital podcast videos, which is not quite in the same direction as Sora," the staffer said.
A number of Sora-concept Chinese stocks hit limit up on Monday, with Easy Click Worldwide, Hangzhou Arcvideo Technology Co Ltd (688039.SH), and Guangdong Insight up more than 20%, and Hanwang and Focus Tech up over 10%.
Dahua rose more than 3%, while Hikvision opened higher and then fell, closing up just 0.45%.
The development of text-to-video capability has raised concerns since December 2023, when Stanford University's Fei-Fei Li team partnered with Google to launch W.A.L.T, a Transformer-based diffusion model for photorealistic video generation.
Sora, a video generation tool that popular chatbot ChatGPT owner OpenAI launched last week, is also a Transformer-based scalable video tool.
Although AI video generation is not new, the launch of Sora is likely to push up the AI multimodal buzz.
China Fortune Securities said in its latest research note that Sora's core technology is based on OpenAI's deep accumulation in natural language processing and image generation, and compared with Runway, Pika, etc., Sora has iconic value in terms of realism and detailed performance of video generation.
The brokerage pointed out that investors can pay attention to AI multimodal applications to shape the new paradigm of digital content production and interaction, empowering the visual industry, from text, 3D generation, animation, movies, pictures, videos, episodes, etc., which is expected to bring about a prosperous development of the content consumption market.
