
Photo 294082396 © Tero Vesalainen | Dreamstime.com
Beneath the digital veneer, a new face-altering tool emerges, leaving critics to wonder: are we opening Pandora’s box? Deep-Live-Cam, a deepfake software recently topping GitHub’s trending page, is stirring up a great deal of attention.
Developed by the team at hacksider, this Python-based tool allows users to perform real-time face swaps and create deepfake videos with just a single image. Its creators tout potential benefits for artists and content creators, empowering them to animate custom characters more efficiently.
Users upload a desired face, and the tool processes frames in real-time, swapping faces and creating an output file. It even offers a ‘Live’ feature for streaming with the fabricated profiles.
Here’s a demo of a speaker turning into Tesla CEO Elon Musk:
Leveraging advanced machine learning models like GFPGAN and inswapper, Deep-Live-Cam achieves high-quality face swaps on both CPU and GPU platforms.
Noting the ethical implications inovlved, the developers say Deep-Live-Cam has built-in checks to prevent the creation of inappropriate content, adding that they’ll add watermarks or shut down the project entirely if legally required to. Users are also advised to ensure consent from people whose likenesses are being utilized.
On X (formerly Twitter), however, users have expressed some grave concerns about how realistic the outputs are. One sardonically comments, “Remember to establish code words with your parents, everyone,” highlighting the growing potential for deception.
Another cuts to the chase: “I can't think of a single application for this technology that isn't fraud or human trafficking-related.” Further, some point out the existence of an ‘NSFW’ option in the tool, despite its creators warning users against creating “inappropriate media including but not limited to nudity, graphic content, [and] sensitive material.”
In this malleable new reality, the face staring back from the screen may no longer be one you truly recognize.
[via Deep-Live-Cam]