AI creates a fake video that a person is talking like

One aspect of artificial intelligence that we have not noticed, is that it can create fake audio and video that is so hard to distinguish. The advent of Photoshop has made us doubt what we see, and what happens when we cannot trust our senses.

The most recent example of image magic - the sound of the AI ​​from the University of Washington, where researchers are creating a new tool that can use audio files and turn them into very smooth lip motions Real, then turn them into videos. The result is a video someone is talking about (something they didn't say). It sounds complicated but you can watch the video below to understand more about this process.

You can see two parallel videos of former US President Barack Obama. The video on the left is the source to get the original sound, and the right video is a completely different talk that the researcher uses algorithms to create and upload video. The 'prepared' video is not really perfect (Mr. Obame's mouth movement is a bit fuzzy - a common mistake with an AI-generated image) but it looks convincing at first.

Researchers say they choose Obama to do the example because the former president's high-quality video is very much, thus making it easier to learn neural networks. Researcher Ira Kemelmacher says they need data of 17 hours of video to follow and copy oral motions, but in the future may be reduced to an hour.

AI creates a fake video that a person is talking like Picture 1
AI controlled video from Obama's audio and visual data

The group behind the study said they hoped to use it to improve video chat tools like Skype. Users can choose the video they speak to teach the software, when it comes to talking to someone, videos can be created automatically and use their voice. This will help in situations where the network connection is not good or want to save mobile data.

Of course, there are concerns that this tool will create fake, misleading videos, fake news. Combining this tool with technology can create anyone's voice with just a few minutes of sample audio. Similar studies are also changing the way of face recognition in real time, creating 3D models of human faces from several pictures .

The research team from Washington University also understands that they will not use it for improper purposes, stating that they only teach Obama's voice and video networks. 'It is impossible to take anyone's voice and turn it into Mr. Obama's video,' said Professor Steve Seitz, 'we will not work to bring what one person says into another's.' But in theory, this technology can create voices on anyone's face.

Watch the video Mr. Obama made by AI created below.