Get Instant Selfie Coaching from Kylie Jenner’s AI Clone

While the generative side of AI is on the rise these days, someone out there is making unconventional tech from radical ideas, and from their list of intriguing innovations is the possibility of "getting selfie advice from a Kylie Jenner voice clone." It is surprising in this era, and it points to more possibilities with AI as well as raises even more questions about its impact on the world.

With a background in electronics and being an artist combined with his experience as a concept developer, Dries Depoorter gets rather weird ideas, good or bad, and tries to turn them into tech you don't normally see around, such as a clock that shows how much of your life is completed, a chat app you can only use when you have less than 5% battery, and AI that tags politicians on social media when they use their phones in livestreams.

To Dries, he only "wanted to make AI unfamiliar to himself again," and so he has been developing this fascinating tech with his one-man team. And spark a lot of curiosity, questions, and interesting ideas in people.

In an interview on Vergecast, he shared the inspiration behind some of his products and how they are being used today. The most intriguing one to me was The Follower, which uses open cameras and AI to find how an Instagram photo is taken.

I had previously heard of something just like The Follower from Mr. Beast, the biggest YouTuber, but only understood how it works better and about Dries' model in The Vergecast episode. Mr. Beast shared about how a kid ran up to him in public, at a store or something, and said that "It works," alluding to a bot that notifies people in realtime of the locations of celebrities based on photos of them shared on Instagram in that same moment. And Mr. Beast had indeed taken a photo with a different fan earlier.

How The Follower works is that it takes photos posted on social media and matches it with footage from open (unsecure) cameras in the location where the photo was taken. This "extra perspective" provides a view of the history you don't see behind the photos. And this model is typically used for celebrities.

Another innovation of his is called Jaywalking, which allows users to catch and report people crossing streets in ways that violate traffic regulations, are unsafe, or are considered careless.

‘Jaywalking’ gives us an opportunity to watch traffic webcams and decide on the fate of pedestrians recklessly crossing the road. Monitors display LIVE unprotected surveillance footage of intersections in different countries. Depoorter then presents us with a dilemma: will we report the unsuspecting jaywalker? A single press of the button can send a screenshot of the violation to the nearest police station with an email that only exists for 10 mins. _Source

This kind of technology would only work in developed countries that have most of their streets equipped with adequate surveillance and traffic systems. And the general outcome I see is improved behavior in people when crossing the streets—not only by being caught when you thought no one was watching, but also by witnessing technology in action and perhaps reporting people yourself with the press of a button.

There are many more that Dries has developed, and they are all on his website, but something very fascinating he created is a selfie coach with the internet personality's voice, Kylie Jenner.

The Selie Coach, which is only experimental for now, takes photos of you from your webcam and directs you in the voice of Kylie Jenner on how to take better selfies.

Using APIs from OpenAI, it analyzes the image and generates a prompt asking for funny advice in the voice of Kylie Jenner, and the API from ElevenLabs generates the voice for the selfie coach with advice just like the following:

“Ok, love the candid vibe, but let’s add some drama. Turn towards the light, lose the headphones, and think mysterious thoughts to spice it up, cuz lighting is everything, babe.”