Technology is becoming more and more advanced and artificial intelligence is even capable of emulating very human gestures.
A group of scientists from Stanford University, the Max Planck Institute for Computer Science, Princeton University and Adobe Research have created a tool called “Editing a Talking Head in Text-Based Video”. They edit the video of a person speaking and, based on the text transcript, they create a realistic video with the modified speaker’s speech.
This type of editing already existed and is being perfected more and more, but the curious thing about this new tool is that, despite how complex it is, it is very simple to use. The research has shown a video of his tool that shows how easy it is to replace the word “napalm” with “French toast” in the phrase from the film Apocalypse Now “I love the smell of napalm in the morning”. In fact, their work was shown to a group of 138 people who, 59.8% of the time, thought these fake videos were real.
To make this artificial intelligence work so easily, the developers have worked on perfecting “phonemes, face posture and geometry, reflectance and expression, and scene lighting per frame.
Despite how advanced it looks, this video edition has limitations. Even so, scientists point to their work as an “important first step” toward a “fully text-based edition”. Researchers are aware that the tool can be used “to falsify personal statements or slander important people,” but they propose certain techniques such as watermarking to make good use of this edition before the tool is publicly launched.