Similar Sounds is a first-of-its-kind feature, powered by audio science and machine learning, that allows musicians to drill down on any sound in the Splice Sounds library and discover sounds that are sonically similar.
With Similar Sounds, Splice is pushing the fields of search and discovery for sound. The conventional approach to categorizing sounds relies on describing the attributes of a sound with tags and metadata — for example, instrument, BPM, genre, and key. But this language-based way of categorization inherently has its limitations and biases. Similar Sounds is built upon on the purest understanding of a sound — the sound itself — and is a groundbreaking application of audio science and machine learning designed to improve a musician’s ability to find the sound they hear in their head.
Similar Sounds uses a combination of unsupervised and supervised machine learning to capture even the slightest, most subtle attributes of sound for which there are no precise words or descriptions. The result is an AI engine that knows how to dynamically capture the most important attributes of each specific sound — from pitch to melody to harmonic profile to temporal structure and even semantics for vocal chops and loops. With this information, the technology can measure how close two given sounds are to each other.
Similar Search doesn’t replace tag-based search — it augments and enhances it. Musicians will now have the ability to navigate the millions of sounds in the Splice Sounds catalog by combining metadata-based search and sonic similarity search, making the process of finding the right sound on Splice both the easiest and most advanced in the market.
We think of Splice as an instrument that musicians play. As a musician, you hear a sound in your head or a vibe you’d like to achieve, and you come to Splice to execute on that inspiration. Or perhaps you just sit down with Splice, as you would sit at a piano, to noodle around to find inspiration. If Sounds is an instrument, Similar Sounds is a fine-tune knob that lets you tap into advanced audio science to find a collection of sounds that are similar to a reference that speaks to you.
November 21, 2019