The specialists taught the neural network to voice image
The program analyzes photographs and paintings, and adds to them the appropriate sounds.
Japanese company Qosmo developed a neural network Imaginary Soundscape, which analyzes pictures and photos and finds them a suitable soundtrack.
For example, if the neural network recognizes the picture shows a dog include a dog barking, and urban landscapes will select the appropriate hum of machines and people talking.
Copes program and abstract paintings, but her greatest manage simple stories, where the accuracy of the selected sounds is nearing 100%.
Just a collection of neural networks about 52 thousand different audio tracks. So choose her.
During the download an error has occurred.