Microsoft has made significant progress in artificial intelligence-based content creation by unveiling a new model capable of creating advanced videos of real people speaking.
The AI-powered VASA-1 model can transform an image into a single video with added audio track.
The company claims that the videos created include lip movements synchronized with the audio, as well as facial expressions and head movements to make them look natural.
Microsoft does not intend to launch any product or API using the VASA-1 model, due to the risk of deepfakes resulting from this technology.
Microsoft has redesigned the operation method of the AI model and highlighted its capabilities. The company claims that VASA-1 can produce videos at a resolution of 512×512 pixels at a rate of up to 40 frames per second.
The AI in generating online videos exhibits almost imperceptible latency. VASA-1 provides up to a minute of high-quality videos using a single still image.
The company emphasized its ability to produce synchronized lip movements with the audio file and facial expressions that match it.
The AI video generation model offers precise user control over various aspects of the video, such as the main gaze direction, head distance, and more.
These factors help in adjusting the 3D head position and facial dynamics, enhancing the ability to customize the output according to the user’s needs.
Additionally, the AI model can also produce diverse video clips using artistic images, singing audio, and non-English speech.
Researchers at Microsoft indicate that the capabilities for these tasks were not present in the initial data, demonstrating the model’s self-learning ability.
The company affirmed that it has no plans to release the AI model to the public, but aims to create interactive virtual characters using it.
Microsoft said: “Given the possibility of misuse, we must recognize the significant positive impacts of our technology, such as supporting educational equality, improving access for individuals facing communication difficulties, and providing care and therapeutic support for those in need.”
The company stated: “We are committed to developing responsible artificial intelligence to enhance human well-being.”
