Sign In

Communications of the ACM

ACM TechNews

Google's AI Invents Sounds Humans Have Never Heard Before


The NSynth project aims to give musicians a new range of tools for making music.

Google researchers have found a way to use artificial intelligence to produce new sounds using the mathematical characteristics of the notes that emerge from different instruments.

Credit: Getty Images

Google researchers, in a project called NSynth, have found a way to use artificial intelligence to produce entirely new sounds using the mathematical characteristics of the notes that emerge from different instruments.

NSynth aims to give musicians an entirely new range of tools for making music.

The researchers work for Google Magenta, where researchers are exploring the limits of neural networks and other forms of machine learning. Magenta is focused on using neural networks to teach machines to make new kinds of music and other art.

NSynth begins with a massive database of sounds, collected from about 1,000 different instruments and entered into a neural network. By analyzing the notes, the neural net learned the audible characteristics of each instrument and created a mathematical "vector" for each. Using these vectors, a machine can mimic the sound of each instrument, as well as combine their sounds.

From Wired
View Full Article

 

Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


 

No entries found