How Does NSynth Super Work as an Instrument?

NSynth Super is an open-source experimental instrument that lets musicians play new sounds generated by the NSynth machine learning algorithm. It works by precomputing more than 100,000 sounds from 16 recorded source sounds, loading them into a physical prototype, and letting a performer choose four source sounds with dials and then sweep a finger across a touchscreen to move between the generated results. It is a research experience prototype from Magenta, a research project within Google, made in collaboration with Google Creative Lab—not a commercial product.

What the instrument actually is

NSynth Super is described on the site as "an experimental physical interface for the NSynth algorithm." The experience prototype was shared with a small community of musicians to understand how they might use it in their creative process. That framing matters: the device exists to explore a way of playing machine-generated sound, not to serve as a finished synth.

The sound pipeline, step by step

1. Source sounds are recorded

For this experiment, 16 original source sounds across a range of 15 pitches were recorded in a studio. These are the raw acoustic inputs the system draws from.

2. NSynth generates new sounds

The recordings were input into the NSynth algorithm to precompute new sounds. The outputs—over 100,000 new sounds—were then loaded into the prototype.

3. Dials select the ingredients

Each dial is assigned 4 source sounds. Using the dials, musicians select the source sounds they want to explore between.

4. The touchscreen navigates the results

Dragging a finger across the touchscreen navigates the new sounds, which combine the acoustic qualities of the 4 source sounds.

Why it sounds different from a sampler or a blend

The site is explicit that NSynth does not combine or blend the original sounds. Instead it synthesizes an entirely new sound using the acoustic qualities of the originals—so, in the site's example, you could get a sound that is part flute and part sitar at once. On NSynth Super, that synthesis happens ahead of time: the 100,000-plus sounds are precomputed and loaded, and the instrument's job is to let you move through them smoothly rather than to generate them live.

What a session looks like in practice

A musician sets the dials to pick which four source sounds to explore between, then drags a finger across the touchscreen to travel through the generated sounds that sit between those sources. The dials define the space; the touchscreen is where you play within it. Because the sounds are precomputed, navigation is the performance gesture—closer to steering through a mapped terrain of timbres than to turning knobs on a traditional synth.

How to get one

The site addresses this directly under "How can I get an NSynth Super?" and describes the instrument as open source. The page does not state a price, a purchase channel, or availability details, so treat those as unconfirmed rather than assuming the instrument is free or orderable. If you want to pursue it, the productive path is to start from the project's own open-source materials rather than from retail expectations.

Who this is for

NSynth Super suits musicians and technologists interested in how machine learning changes sound design and performance—people who want to feel out an interface for navigating generated timbres. It is less suited to anyone seeking a finished, supported instrument with documented pricing and distribution; the site presents it as an ongoing experiment, and the prototype reached only a small community of musicians.