On the social network Facebook, the developers made an algorithm based on neural networks. He knows how to correct failed photos in which the character closed his eyes or accidentally blinked. To open them, the algorithm of the unsuccessful photo and a few “successful” shots is enough. The neural network Generative Adversarial Network (GAN) has been trained by developers not only to recognize closed eyes in the photo. He can replace them with suitable open ones and even retouch what happens. During operation, it takes into account all parameters: lighting, foreshortening, makeup, and more. From left to right: a successful shot of the object, the initial failed photo, the result of processing in Photoshop, the operation of the algorithm from Facebook Often, algorithms of this kind leave much to be desired. Artificial intelligence may not understand things important to humans. Say, opening your eyes, do not change the color of the skin nearby. Or that you should not add too many cilia. But Facebook’s development is superior. The authors of the project illustrated this with photographs: unsuccessful, with open eyes for processing, after working in Photoshop and one after editing with a new algorithm. The brainchild of Facebook has achieved a much more natural result than Photoshop. It does not seem that the eyes are cut from a different picture, they do not show either a mismatch of colors or noticeable cross-linking.