Google NSynth Super

Working with Google to develop their groundbreaking machine learning software.

Following an extensive project working alongside the team at Google – we're delighted to announce our involvement in the Google NSynth Super.

The NSynth Super uses machine learning to combine the sound of pre-existing instruments, allowing users to merge sounds to an extent which wasn’t possible before.

Working alongside the team at Google. we brought a whole range of instruments into our studios for our Sound team to record – including a Sitar, Marimba, Tubular Bells, and Crotales amongst many others, and even visited a motorcycle repair workshop to record the sound of a vintage Norton Commando Production Racer, also known as ‘The Yellow Peril’.

The recordings were then imported into an algorithm and turned into code, allowing them to be modified in a way a usual sample couldn’t be.

Overall, Google’s algorithm contains over 300,000 instrument sounds, making the open source NSynth Super the largest audio dataset publicly available.

The ability to develop such unique sounds allows musicians further freedom in their music making – deconstructing many rules and expectations of their compositions, and making room for new, unpredictable results from the Nsynth Super.