D-id – Deep Fake in White

The post was originally published in Russian on Startup of the Day. Alexander kindly agreed to republish what we think is of great value to our readers.

I used to write about D-id, the Israeli startup of the day, five years ago, it had just graduated from Y Combinator at the time. The startup wanted to apply filters on face photographs in a way that the person didn’t notice the difference, and recognition algorithms malfunctioned. The goal is to prevent ill-intended people from tricking biometric authorization systems with public photos.

One year after the accelerator it indeed released the product, then there were a couple of rounds more, and this March it brought in USD 25M with a completely different concept. There’s not a trace of old positioning on the website now, D-id is doing something almost opposite now.

Now, its neural networks take a person’s photograph and voice and use them to make a video with the desired movement or speech – we would call it deep fake if it were operated by an enemy hacker and not a respected B2B startup. With D-id’s help, a TV broadcasting company can do without a host, the university can do without a lecturer, and a corporation can send out a personal address from its CEO to every employee. And all one needs for that, other than a voice recording, is just one photograph!

‘Resurrection’ of old photos from archives is a separate direction, D-id boasts about it almost as much as commercial generation. To my taste, the examples look disastrous, I wouldn’t like to turn my great-grandmother’s portrait into this, but some people like it. The startup even partnered up with MyHeritage, a large online project on genealogy.

https://www.d-id.com/

#tech #roundb #israel #ai