How this Chinese multinational technology company, Baidu’s AI produced news videos using only a URL? Feeding just an article URL it produces lifelike news videos.
AI is one of the ways of which contradictory views have been developed in news output. Firstly, publishing companies might generate more content in a less effortful manner, secondly, that might strip away the emotional aspect in broadcasting or keep employees out of jobs.
But how Baidu’s AI produced news videos?
The buzz around major Western technology companies such as Amazon or Google tends when thinking of AI-powerful.
UK Alphabet subsidiary DeepMind has made huge strides with its AlphaGo Master Go player Lee Sedol, for example, in the best of five games, which is much more complex than chess.
However, the fact that giants such as Alibaba and Tencent play immensely in both the AI and the broader technology scene cannot minimize Chinese corporations’ size.
But Baidu’s AI produced news video, such news is not so sudden while Baidu is another example that runs one of the most popular search engines worldwide and acts as a Chinese counterpart to Google in several specific ways.
In 2018, the Chinese news service Xinhua made its debut with an AI anchor. The agency released an improved version of human voices and actions earlier this month. AI has advanced in text-based news in algorithms that compose fantastic headlines. The search giant in China, Baidu developed Vidpress, a new AI model that combines video and text with articles.
Vidpress has now been deployed by the company in its Haokan short video app and works only with mandarin. The AI algorithm can produce up to 1,000 videos daily, much more than the 300-500 currently released by the human editors. In 2 and half minutes, Vidpress can produce a two-minute 720p video, whilst human publishers use an average of 15 minutes to do this.
Baidu used thousands of articles online to develop this model to understand the context of a news story. Besides, AI models for voice and video generation had to be trained individually. The algorithm synchronizes both streams for a smooth final video in the final step.
When you fill a URL in the AI algorithm, all related articles are automatically retrieved from the web and summarised. For example, if your input is a story about Apple’s new iPhone launch, the AI will retrieve all the details of the launch including price and specifications. You will now search for images and clips in your media library and on the web to create a video.
Baidu said that its AI also can detect social media trends and create videos on related issues. The company says that Baidu aims to provide news agencies and creators with videos with synthesized stories.
The search giant said that the AI looks the most trustworthy sources to build material, but does not mention the sources. As the company now operates this model in a controlled environment, it may have to put some guards when they are released for more people. Moreover, publishers must also be assigned to check whether someone produces videos that spread or violate copyrights.
Overall it hasn’t been clear which video has been produced. No link to the video has made public yet! If such turns out to be very humanly then I must say it will have a serious impact on journalism. Further, any link to the video might or might not be published as all such things are under testing. Commercialised produces may be conducted by the company in the future.