A new, freely available AI deepfake tool lets you use a single photo of anyone’s face to imitate them online in a video call.
The open-source software tool is called “Deep-Live-Cam” and has gained a lot of attention online thanks to its ease of use and interesting ability to create a deceptive persona through video so simply.
Recently, an X user by the name of Joao, showed viewers just how effective and easy the software tool is to use.
In a post on X, he demonstrated himself working with Deep-Live-Cam to realistically superimpose the faces of Meta founder Mark Zuckerberg, George Clooney, Hugh Grant and Republican Vice Presidential nominee JD Vance over his own during a live video.
Real-time deepfakes are here. I used the repo below to swap my face directly in my webcam feed as JD Vance, Hugh Grant, Zuckerburg, and George Clooney.
We know people have been using this kind of tech for malicious purposes… Don’t trust your eyes, don’t trust your ears. pic.twitter.com/iWvgtwtXLD
— joao (@jay_wooow) August 9, 2024
The project has been under development since late 2023, so it’s pretty new, but its more recent popularity is due to a surge in videos on social media showcasing how it works.
The open-source project behind Deep-Live-Cam even took the top spot on GitHub’s trending repositories list thanks to public curiosity, according to ArsTechnica.
With Deep-Live-Cam (which probably could have more fittingly been named Deep-Lie-Cam), users can take any one photo of a person’s face and instantly have it overlay their own face during live webcam video feeds.
The software then takes the characteristics of the original face and uses them to replicate complex poses, expressions and so forth, even with realistic adjustment for lighting in real-time.
Obviously, since this is a type of generative AI we’re talking about here, there are occasional flaws in the rendering, but the software is remarkably advanced despite that, and tools like this will only get better, sooner rather than later.
Also, keep in mind that this is just a single open-source tool. A more robustly developed software system with more resources backing it could probably achieve much better results after some development effort.
Joao’s video on X is especially striking because it shows just how well the software manages to let him replicate such famous faces so easily.
Another researcher, Matthew Berman, also posted his own video to X showing himself imitating X owner Elon Musk in a real-time webcam video.
#1 trending github repo right now looks INSANE
Single image to live stream deep fake.
Look at that quality!!
It’s called Deep-Live-Cam (link in replies) pic.twitter.com/VgtnJ1LKRM
— MatthewBerman (@MatthewBerman) August 9, 2024
Deep-Live-Cam uses a pre-trained AI model called “inswapper” to switch faces during videos and then further refine the results for artifact removal with yet another AI model called GFPGAN.
The first of these tools, inswapper, is no lightweight piece of AI tech. It has been trained by its creators with a dataset containing millions of human facial images from thousands of people.
This face database included diverse lighting conditions and facial changes to further perfect inswapper’s training.
As is the case with all deepfake technology, the potential for fraud and dishonest use of Deep-Live-Cam is just below the surface, though the tool itself and others like it are examples of software that’s generally legal in most places.
“Hacksider”, the creator of Deep-live-Cam states in a disclaimer on its Github repo page,
“This software is meant to be a productive contribution to the rapidly growing AI-generated media industry. It will help artists with tasks such as animating a custom character or using the character as a model for clothing etc.
The developers of this software are aware of its possible unethical applications and are committed to take preventative measures against them. It has a built-in check which prevents the program from working on inappropriate media including but not limited to nudity, graphic content, sensitive material such as war footage etc.”
If you’d like to explore Deep-Live-Cam from its source, bear in mind that we can’t vouch for the safety of any files for this or any other software in Github.
Credit : Source Post
