NSynth Super is an experimental physical instrument built around NSynth, the Neural Synthesizer, a machine learning algorithm from Magenta, a research project within Google. It was created by Magenta in collaboration with Google Creative Lab and is presented as an experience prototype that was shared with a small community of musicians.
NSynth uses a deep neural network to learn the characteristics of sounds and then synthesizes entirely new sounds at the level of individual samples, rather than blending or layering the originals. The result can be a sound that is part flute and part sitar. NSynth Super makes this playable. Sixteen original source sounds were recorded in a studio across 15 pitches and run through the algorithm, producing more than 100,000 new sounds that are loaded onto the device. Each of the instrument's dials is assigned four source sounds, and musicians pick the sounds they want to explore between, then drag a finger across the touchscreen to move among the combinations. The instrument can be played from any MIDI source, such as a DAW, sequencer or keyboard.
It is aimed at musicians, makers and researchers interested in new ways to use machine learning in sound design. The hardware and software are open source, and the project is described as an open source experimental instrument. The site lists no price and does not sell a finished product.