Jukedeck is an AI that incorporates the use of deep neural networks (DNN) to compose original soundtracks from scratch. The DNN is used to learn the music composition at a granular level.
Wait a minute. How does it all happen?
The procedure of composing each soundtrack is divided into two parts - composition and synthesis.
In composition phase, a symbolic representation of the music is generated. Similar to traditional music, the symbolic score is a combination of notes and chords. Synthesis phase, on the other hand, converts the generated symbolic score into music.
The composition of music is done with the help of DNN. It selects which notes and chords are to be played at a particular time.
For the synthesis, notes composed by the DNN are then converted to music by AI powered audio synthesis algorithms developed by the team at Jukedeck.
In the example below, composition system was asked for 8 bars of piano music to synthesise these 8 bars using an upright piano sound.
Awesome! Isn't it?
This was the first time that a computer has written and produced an entire soundtrack, purely using machine learning techniques [only duration and genre was specified as an input to the AI initially].
This work represents a significant breakthrough for researchers working on developing intelligent 'creative' machines.
If you like the post, please upvote, comment or resteem!