Awesome/Creepy AI Brings Historical Pictures To Life And Makes Them Move


Katy Evans

Katy is Managing Editor at IFLScience where she oversees editorial content from News articles to Features, and even occasionally writes some.

Managing Editor

Marie Sk?odowska Curie

Marie Sk?odowska Curie, staring through time and finding us wanting. Image credit: MyHeritage

A new AI-generated photo service called “Deep Nostalgia™” allows you to turn static photos into Harry Potter-esque moving pictures that look both awesome and, frankly, a little creepy.  

The service, offered by the online genealogy site My Heritage, was designed for people wanting to animate the faces of loved ones in family photos – perhaps those who have died or ancestors they’ve never met. “Experience your family history like never before!” the site proclaims. 


This is quite a sweet idea and many people have been brought to tears, sharing their stories on social media of seeing lost loved ones or “meeting” family members for the first time.  

Of course, it couldn’t stay wholesome and some people have run with it, not just recreating moving pictures of famous faces – Marie Curie, Amelia Earheart, Rosalind Franklin have proven popular – but pushing the AI to the limit to see if it works for objects like statues or paintings too. Alas, it does, so if you’ve ever wanted to be stared at by the blank-eyed statue of Michelangelo’s David or Tutankhamun's death mask come to life, nows your chance.  


Developed by technology company D-ID, Deep Nostalgia™ uses pre-recorded videos of facial movements. It digitally recreates naturalistic movements using algorithms to identify which ones work best for each photo, depending on the subject’s pose.  


You just upload a photo (you have to sign up for a free MyHeritage account) and it will work its magic as you watch, applying subtle facial movements – eyes narrowing, a head turning, an eyebrow raising – creating a moving image that, for a few seconds, appears more like video footage. 


Most “deep fake” videos use generative adversarial networks, or GANs, to create original footage. This works by using two “adversarial” AIs, one to create the content, the other to evaluate how close to the real thing it is. The two AIs constantly learn from each other, pushing them to get better and create the most convincing fake video. 

However, Deep Nostalgia doesn’t use audio and only lasts for a few seconds in order to avoid people creating deep fakes, MyHeritage said.  


It has its limitations too. It only works on faces at the moment so there’s no chance of recreating people getting up and walking around, including ancient Egyptian mummies. It also struggles slightly with filling in missing features like teeth or ears, or when glasses partially obscure the face.  

However, depending on the quality of the photo uploaded, it can also be scarily good. Which, of course, meant that people on social media have been having some fun.