Russian tech giant Yandex has taught its voice assistant Alisa to draw pictures at user command, a spokesperson with the company told TASS.
“Alisa now knows how to create pictures at a users’ request. Just ask her and she will draw anything, whether it is an olive grove or a deserted beach. This is the first voice assistant in the world that generates pictures on any topic,” Yandex says.
The “artist” feature is available on the Yandex mobile application and in Yandex.Browser, where the Alisa voice assistant lives.
“In order to get a picture from Alisa it is enough to ask her: Alisa, draw a panda or Alisa, draw an autumn landscape. She can even depict something that is hard to imagine — for example, inspiration. It could be that this sort of picture will turn out to be a real masterpiece,” the company explains.
The voice assistant creates images on demand using two neural networks. One of them was trained on the works of artists of the 20th century and knows how to generate pictures, while the other neural network determines what is depicted on them. The developers used search queries to teach the neural network to compare the semantic proximity of images and words.
Besides that, Alisa can guide you on the works of famous masters. The voice assistant also has the ability to recognize images on pictures and tell a person about his or her resemblance to some famous people.
Such a system is from Google and NVIDIA, although no one has yet thought to combine it with the voice assistant. Overall, the images look pretty creepy, which is understandable — they drew people.