DataMind Audio's Combobulator resynthesises audio through neural networks to recreate other artists' styles
Artificial intelligence is fast revolutionising music software; recently, Google and Adobe both announced the next steps in their experimental adventures into music AI.
The latest development in this sphere is one of the most intriguing AI-based plugins we've seen in recent months. Combobulator from DataMind Audio processes incoming audio through neural networks trained on the music of artists and sound designers, resynthesising the audio using a process called "style transfer" to recreate the style of another artist using your sounds.
DataMind says that this lets you play "another artist's brain like an instrument" by synthesising new audio based on a combination of the input signal and the information in the neural networks. In practice, Combobulator acts more like an effects plugin that applies a new timbre to a sound while preserving its rhythmic and melodic content.
To start, you can select which neural network, or Artist Brain, you want to use. The plugin arrives with two Artist Brains, but more can be downloaded through DataMind's Model Marketplace. Each Artist Brain is developed in-house by musical AI experts and 50% of sales go to contributing artists.
"We're trying to set up a new standard and a new cottage industry for music producers to make income on AI that's trained on their likeness," explains DataMind's Ben Cantil. "We like to think we're doing farm-to-table AI."
At the centre of the plugin's interface is a visualiser with two sliders around the outside to control Scale and Offset. Both of these parameters control "how the neural network thinks";
