New Chinese AI video generator unveiled to rival OpenAI's Sora

CGTN Europe

A new AI video generator called Vidu has been unveiled at Tsinghua University in China. It’s being titled as a rival to OpenAI’s Sora system and is one of many new AI tools emerging in China.  #ai #china #vidu #sora #openai

Transcript

00:00A new Artificial Intelligence video generator has been unveiled by Tsinghua University in China.

00:07Known as VIDU, it's been touted as a rival to OpenAI's Sora system.

00:12It's one of many new Artificial Intelligence tools emerging in China.

00:17Our correspondent Jen Copestake reports.

00:20Large-scale AI models are creating new ways of working.

00:24Powerful tools for data analysts can help companies create content with the aim of grabbing audience share.

00:30In China, public opinion platform GS Data is using AI to analyze nearly 90 days of online data

00:37to identify the most popular topics among its users and forecast future behavior patterns.

00:44It says it knows how users feel about content,

00:47with a sentiment index claiming a 70-80% success rate of predicting emotion.

00:53Behind this, we have a strong team of skilled AI algorithm engineers providing support.

00:59First and foremost, we require massive data sets that are meticulously labeled and processed.

01:05With accurate labels, our unique algorithms enable us to analyze this extensive data

01:10and derive the results we desire.

01:13Beijing's Tsinghua University recently unveiled VIDU,

01:17a video generation competitor to OpenAI's Sora system.

01:21VIDU utilizes our proprietary technology architecture.

01:25Sora does not disclose its technical roadmap.

01:28We have been developing core technologies such as deep artificial intelligence and diffusion models independently.

01:35VIDU creates 16-second long HD videos with a single click

01:40and can understand and generate content specifically aimed at Chinese users.

01:45For example, the dust billowing up we see as the car moves

01:49and the sunlight and shadow effects at different times of the day can be rendered quite realistically.

01:55Another feature is it can comprehend some of the language of multiple camera angle utilization

02:01and focusing and light tracking effects, even some illusory scenes you can envision.

02:08Such large language models help content creators save time in editing

02:12by understanding different formats and recommending music.

02:15And this technology can even replicate humans in creating digital anchors,

02:21mimicking individual characteristics like voice, lip movement, and lifelike facial features.

02:27The generation of digital people is actually very simple, roughly divided into two steps.

02:32The first step is to clone your image. The second step is to clone your voice.

02:37We only need to collect some of your video footage.

02:40Users of digital human products can use text to generate new content,

02:44allowing digital characters to speak in multiple languages.

02:48These developments come with ethical challenges and regulatory issues.

02:52One example is reviving dead celebrities or swapping the faces of well-known people.

02:57Balancing innovation with necessary compliance will be critical as these new technologies expand rapidly.

03:04Jen Copesteak, CGTN.

Category

Transcript

Recommended