An AI capable of mixing musical genres and instruments to create music from A to Z.
Jukebox is an artificial intelligence system based on machine learning, capable of mixing musical genres and instruments to create music from scratch. You will discover that the latest creations of OpenAI’s artificial intelligence look strangely like the sounds of King Elvis Presley..
Jukebox can create a sound from A to Z
A year ago, OpenAI introduced MuseNet. Another artificial intelligence capable of generating sounds. An intelligent system capable of en paying attention to a particular music for long minutes, in order to understand it perfectly. Ai can therefore understand the general context of a song’s melodies, rather than how they follow each other in a short section. Goal: to predict the rest of the song before she has even heard it. .
Jukebox is an improved version of MuseNet, another AI from OpenAI. More than predicting the sequel to music, this new artificial intelligence is capable of creating sounds from A to Z. A bit like Amadeus Code, a composer IA created by Jun Inoue, Gyo Kitagawa and Taishi Fukuyama.
en The solution devised by OpenAI is able to compose a song by choosing the right instruments, mixing styles and defining a text with more or less clear lyrics… We’ll let you listen:
A powerful but useless AI?
To achieve this feat,Jukebox breaks down existing music into different “pieces” and then reconstructs a final track. This is the heart of OpenAI’s new research. Computers do not have the ability to memorize text. If this task seems simple because it is innate in humans, machines need advanced learning to achieve this..
So ai has to learn to choose a voice, understand what it says and harmonize tone, tempo and everything else. Jukebox is one of the few artificial intelligences artificielles in the world to be capable of such a feat.. Formed from 1.2 million songs,the system has several facets, but improvisation remains the main mission of this AI.
Jukebox can use the voice of Elvis Presley or Eminem and make them say anything. Like MuseNet,Jukebox is also able to assimilate the first 12 seconds of a song to generate the sequel in a similar style. Despite great capabilities, Jukebox designers say that:
“We shared Jukebox with an ensemble of 10 musicians of different genres… unfortunately these musicians did not find that our invention was immediately applicable to their creative process.”