Musical composition

Musical composition

This week we take a look at musical composition using AI.

For the exercise, I have used an application called AIVA.

AIVA describes its product as a music generational assistant that can generate new songs in more than 250 different styles in a matter of seconds.

It is possible to create a style model or upload an audio or MIDI file for inspiration. The final product can then be downloaded - either as an audio file or as a MIDI file which allows for further editing.

The popular music streaming platforms are now full of AI-generated music - but is it any good?

For the experiment, I will take one of my favourite soundtracks - The Kings Speech and the title track as the "inspiration" for the AI-generated piece.

audio-thumbnail
The Kings Speech Extract
0:00
/95.13795918367347

I simply selected this piece as inspiration and AIVA and set up a few parameters to generate some brand new compositions.

As you can see, I can choose the instruments that are used, the key signature, the emotion, duration and the number of compositions.

And here are a couple of the results.

audio-thumbnail
The Tims Speech Soft Piano Variation
0:00
/42.370612244897956
audio-thumbnail
The Tims Speech Orchestral Variation
0:00
/56.52897959183674

I feel you can definitely hear the influence of Alexandre Desplat's King Speech within these tracks. Of course, they are vastly inferior to the original, but I am sure they can be substantially improved through tinkering with the settings and further editing.

I don't think we are replacing human composers just yet, but as an idea creation tool pretty amazing. And who knows what will be possible in the years ahead.