#科技头条#【说出来你可能不信,视频里的我可能是假的】

#科技头条#【说出来你可能不信,视频里的我可能是假的】牛津大学的研究人员已经用人工智能做假视频了!只需要人脸的静态图像和音频片段,系统就能识别面部特征,操纵嘴型来和每个词匹配,当播放音频时,看起来就像是你在说话。研究人员称,该技术还可以在说话时,实时制作假视频。随着技术的改进,未来法庭上的视频证据也越来越真假难辨了。

Artificial intelligence can use still images of people and turn them into FAKE video - and even put words in their mouths

  • The system takes an image of a person as well as an audio clip to create a video 
  • It identifies facial features in photos using algorithms that recognize the face
  • As audio plays, it manipulates the mouth so it looks like the person is speaking
  • With improvement, the researchers say the AI could make fake videos seem real
  • The system could eventually render video evidence unreliable in court cases

Researchers have developed an artificial intelligence system that can put words right into people's mouths.

It takes an image of a person as well as an audio clip, using them to create a video of a person speaking that audio.

While the system is still rough and not realistic looking, the researchers claim the software could soon make fake videos that seem real.

As an audio clip plays, the AI manipulates the mouth to look like the person is speaking. Left is the original still image. Middle: Output frames without a skip connection. Right: Output frames with a skip connection. The skip connection helps carry facial features  from the original image 

As an audio clip plays, the AI manipulates the mouth to look like the person is speaking. Left is the original still image. Middle: Output frames without a skip connection. Right: Output frames with a skip connection. The skip connection helps carry facial features from the original image

He said that eventually, audio from news videos could be automatically translated into a different language and the mouth movements updated to fit.

However, this isn't the the only system that can adjust images to different audio.

Other systems have needed a lot of video to work, and they would match up the way a person's mouth moved when they made sounds and then use specific parts of the image in edited footage with new audio.

 

'With the old method, if you’re trying to redub footage of David Cameron, for example, it’s easy because there are hours of video,' says Chung.

However, for less well-known people this would be more difficult, as there aren't many images available to take from.

Chung also thinks the new method his team has developed could be useful for dubbing animations.

 While current video doctoring systems focus on changing the shape of the mouth, Alex Champandard, the co-founder of creative.ai which makes AI tools for artists, says these tools will also be able to change facial expressions and posture in the future

While current video doctoring systems focus on changing the shape of the mouth, Alex Champandard, the co-founder of creative.ai which makes AI tools for artists, says these tools will also be able to change facial expressions and posture in the future

While these systems currently focus on changing the shape of the mouth, Alex Champandard, the co-founder of creative.ai which makes AI tools for artists, says these tools will also be able to change facial expressions and posture in the future.

He says that soon, people will have trouble figuring out which videos are fake and which are real.

Already, given enough time, experts can create fake videos that look realistic.

This is because AI tools are making the process rapid and simple, so much so that eventually, almost anyone could make these videos too.

This could render video evidence unreliable in court cases and could mean that courts will rely more on forensic, scientific evidence and witness testimonies in cases.

Other systems have needed a lot of video to work. This is easy when someone is well known and has many hours of video footage of them available - for example President Donald Trump. However, for less well-known people, this would be more difficult

Other systems have needed a lot of video to work. This is easy when someone is well known and has many hours of video footage of them available - for example President Donald Trump. However, for less well-known people, this would be more difficult

HOW IT WORKS

The system, developed by researchers at the University of Oxford, identifies facial features in photos using algorithms that recognize the face.

Then, as a chosen audio clip plays, the AI system manipulates the mouth in the still photo so that it looks like the person is speaking.

'The application we’re thinking of is redubbing a video into another language,' Joon Son Chung, a graduate student at the University of Oxford and one of the creators of the system, told New Scientist.

http://www.dailymail.co.uk/sciencetech/article-4523714/AI-use-images-turn-FAKE-videos.html


Comments are closed.



无觅相关文章插件