AI for Music Discovery

To understand the potential applications of AI in the music discovery process, a literature study was conducted which resulted in discovering the AI music categorisation tool Cyanite AI¹.

To understand the possibilities of using this tool, available product analysis and tinkering was done, followed by expert interviews with the CEO of Cyanite AI² and a digital designerwho has used the tool to create a music visualisation application³.

Overall, AI has various positive applications in the music industry like music creation, search & recommendation and auto-tagging⁴. Such applications could be used to, for instance, automatically filter a large music catalogue, allowing filmmakers faster access to suitable music.

In practice, to find music efficiently, a well-categorised music library is needed to deliver the tracks that exactly correspond to a search request. Each song is usually tagged with metadata parameters such as the name, artist, genre, mood, etc. The problem is that tagging music manually is one of the most tedious and subjective tasks in the music industry.⁵ The mood evoked by a song can be determined only after listening to it. While doing this for a single song might be manageable, dealing with big music libraries becomes challenging. Moreover, tagging requires extreme accuracy and precision. Inconsistent and wrong manual tagging leads to a poor search experience, which makes it harder to find suitable music. Then, finding music that fits the moment perfectly is like trying to find a needle in a haystack.

However, tagging music is a task that can be done with the help of AI which can be used as a powerful objective tool to tag music at scale. AI-based tagging can increase the searchability of a music catalogue with little to no effort and therefore significantly reduce the need for manual tagging.

The most important discovery was the tool Cyanite AI - an AI music search and tagging engine that can quickly listen to and categorise millions of songs. After uploading the music, Cyanite AI automatically analyses, sorts, compares and structures the catalogue and visualises results. The visual below shows an example of a song analysis from Cyanite AI:

Cyanite AI

The tool has a web application and an API which allows an easy business-to-business (B2B) integration of the tool’s capabilities into any product. In the visuals below, you can see what the web application and API look like:

Cyanite AI web application

Cyanite AI API

Next to music analysis and visualisation, Cyanite AI offers powerful out-of-the-box features to search for music within a catalogue:

Search by audio

Finds music based on a reference song.

Search by keywords

Finds music based on combining and weighting up to 1,500 keywords.

Search by text

Find music based on any given text input.

Those discovered features also matched the search techniques that were mentioned in the description of the chosen idea in the Ideation phase:

1. The search by audio feature perfectly fit the AI similarity searching technique.

2. The search by keywords feature acted as the emotional knob searching technique.

3. The search by text feature had potential to be used as the both the film/video structure and AI similarity searching techniques.

This discovery was later used in the Concept Creation phase to show how these three features of the Cyanite AI tool were incorporated into the concept for the advanced music discovery tool for filmmakers.

The expert interview with Cyanite AI’s CEO provided further explanation about using those features. The expert was interested in the research done so far and provided full access to the tool since the free version has limited functionality of the web application and no access to the API.

The expert interview with the digital designer provided knowledge about how to use Cyanite AI’s API to create custom applications.

The full research material is referenced below.

As AI can speed up the music discovery process and Cyanite AI offers powerful searching techniques (search by audio, search by keywords and search by text) which can be easily integrated to a custom application using the API, it was decided to use Cyanite AI as the core of the solution. This is further explained in the Concept Creation phase.