Microsoft researchers have unveiled their latest AI-powered innovation: creating a video of a person speaking to a camera from a simple photograph. These faces appear incredibly lifelike in terms of replicating the emotions and expressions once animated. Lips, eyebrows and other elements modeled on the voiceover move so realistically that we understand the concerns about the deep fake phenomenon that such technology could create.
Realistic animation of portraits
The platform, called VASA-1, developed as part of this project is capable of generating high-definition video at a frame rate of 40 frames per second. To develop VASA-1, the Mountain View team used several sophisticated technologies related to deep learning. The latter works for animating portraits: by providing it with a photograph and a sound track, the AI can create a video.
This example, taken from a presentation page on Microsoft’s website, illustrates just how powerful (and alarming) the technology under development is. To view all videos, we invite you to visit the dedicated page on the Microsoft website.