Similarity search algorithms usually work to compute the distance between songs in the space of different audio features. Calculating similarity distance allows to input a query that will then return tracks with the shortest similarity distances which means they are the most similar ones.
Similarity search becomes particularly important as music libraries grow and music search and discovery becomes a chore. Music recommendation systems are also based on similarity search algorithms. However, there is a clear distinction between systems that recommend music that would be liked and recommendation systems that calculate similarity based on audio features.
1. Finding similar songs using audio references for sync and music briefs
Music briefs often come with very tight deadlines. In fact, our own research alongside the University of Utrecht has shown that 3 out of 4 music searches are done on a ticking clock. Usually, music supervisors use genre, mood, instruments, and character keywords to find music for sync. So correct metadata is essential. With Similarity Search, it is possible to paste a reference track that accompanies the brief and get a list of similar-sounding tracks.
To identify the right music for the right project, AI that finds similar songs is indispensable as it speeds up the process and increases the chances of licensing every single song in the catalog, no matter how old or niche they are. The big advantage over Spotify’s “Similar Artist” feature: Similarity Search only looks at the actual sound and character of the song to find matches. This proves way handier for use cases where the exact tonality of the song is important.
Photo at Unsplash @dillonjshook
2. Finding duplicates
If you own the music library that you’ve had for a while, there is a chance that there are duplicate songs in there. Removing these duplicate songs considering the size of the library can be a challenge. However, with similar songs finder, you can easily find duplicates in your catalog and it takes no time to remove them. As Similarity Search returns a list of similar songs, all duplicates of the reference track will be listed on top of the list, so they can be removed quickly and easily.
3. Targeting social media campaigns
AI that finds similar songs can be particularly useful to users who are dealing with the development of an up-and-coming artist. Catalog users can use it to find popular artists similar to the breaking artists and then apply this information to set up custom audiences on Facebook, Instagram, and Google.
For example, an A&R or artist manager planning a marketing campaign for a new protege can use the Similarity Search to detect similar-sounding songs from well-known artists and then target their fans on social media. These people have shown that they most likely enjoy this specific sound and will form the core of the audience for the new artist. We described this use case in detail in the article How to Create Custom Audiences for Pre-Release Music Campaigns in Facebook, Instagram, and Google.
Photo at Unsplash @William White
4. Determining type beats
Type beats are instrumentals that emulate the style of a popular artist. Beat producers use them to get more exposure online and piggyback on the fame of their “typed” artist. This practice of emulating someone’s music style has produced a whole economy where familiar beats are sought after.
Algorithms that find similar songs help beat producers validate their beats against artists whose style they are trying to reference. Conversely, Similarity Search is a valuable tool for users using a reference track from a popular artist to uncover the best matching type beats. As the beat space is already overly saturated with type beats similar to well-known artists, the Similarity Search feature will attract beatmakers or catalog users who are interested in identifying matches with up-and-coming or more niche artists to avoid the competition.
5. Playlist pitching
Musicians and labels often use Similarity Search to pitch to Spotify editors or third-party playlist curators. You can ingest full playlists into the Similarity Search and then compare your track to them to detect the closest match – that will be the first target for pitching. This targeted action is the counter approach to spraying and praying the song to every curator leaving most of them annoyed.
Furthermore, Similarity Search can be used to provide familiar references to the editor to understand the profile of the song in the description, for example, “For the fans of Max Richter and Dustin O’Halloran.” This process makes the track “Spotify-ready” or optimized for Spotify curators and various moods and occasions that Spotify playlists represent. Read more about this process here – How to Write Press Releases and Music Pitches with Cyanite. When a catalog offers a Similarity Search function, it automatically offers its users a way to improve their playlist pitching strategy.
6. Playlist optimization
Similarity Search is usually an indispensable part of sophisticated recommendation algorithms in streaming services such as Spotify or Youtube. Essentially, Similarity Search could generate a whole playlist automatically based on a reference track without any human involvement needed. In action, it mostly serves as inspiration for playlist curators to fill playlists with songs that fit.
This combined approach to playlist optimization allows playlist curators to create playlists for a certain environment, for example, study or based on a mood.
7. Dj mixing and DJ Crates optimization
For DJs, features and tools that support e.g. mixing in key and creating better sounding DJ crates are paramount when working with a music catalog. Sometimes a DJ falls in love with two tracks and wants to play both but can’t really figure out a connecting piece to get from one song to the other. They can utilize music search similarity tools to find songs that go along the same lines and provide a smoother transition. The track suggestions can then be filtered out using Camelot Wheel. As a result, this feature allows the DJs to mix music so it matches in vibe and feel.
This use case is described in detail in our How to Use Cyanite to Optimize Your Playlists and DJ Sets for Harmonic Mixing and Similarity article.
Photo at Unsplash @Daniel Eliashevskyi
8. Using own songs and finding similar songs on Spotify to detect blind spots in a catalog
Blind spots in the catalog are the pain of music catalog owners. Old music and niche songs are often stuck in the tail of the catalog with no chance for monetization. Audio similarity search is used to uncover these hidden layers of music. The music search returns up to 500 similar tracks which are a combination of popular songs and songs with undiscovered potential.
The user can then explore which music fits their needs without having to choose only from the known material. In essence, Similarity Search expands the variety and choice of music for users and can help maintain and improve catalog user retention rates.
9. Finding samples across the catalog
In sample catalogs, Similarity Search algorithms can be used to find similar samples. Instead of spending hours digging through sample libraries, the user can select a reference sample or the first sample they like and get a list of similar-sounding samples. They can then refine the search by key or bpm. An analogous technology was used in Splice’s CoSo app, that gives you a place to start with a stack of up to 8 samples which you can then change and adjust by replacing initial samples with similar ones or adding a completely new sample to the stack.
Similarity Search has many applications. Apart from using a reference song to find similar tracks, Similarity Search can dramatically improve the quality of the catalog by removing duplicates and allowing the users to find old and niche songs. Similarity Search is also indispensable in playlist curation and optimization. Less apparent use cases include pitching playlists using similarity results for Spotify descriptions and targeting social media campaigns. As music platforms and social media are an integral part in promoting artists’ work, these latter capacities will become more emphasized and requested by professional musicians.
At Cyanite, we provide the Similarity Search function through an API as well as the web app. The Similarity Search takes a reference track and gives you a list of songs that match by pulling audio, metadata, and other relevant information from audio files. As a result, Similarity Search can reduce search time up to 86% and offload some labor-intensive tasks from users as well as music catalog owners. We show a real example of how similarity search works in a production music library in this video on How Cinephonix integrated AI search into their music library.
Cyanite Similarity Search interface
I want to Use AI to Find Similar Songs – how can I get started?
If you want to get the first grip on Cyanite’s technology, you can also register for our free web app to analyze music and try similarity searches without any coding needed.
More Cyanite content on AI and music
KEY TAKE-AWAYS FROM MUSEXPO 2023 IN LOS ANGELES – PART 1
Written by our CMO Jakob Höflich I just came back to Berlin after visiting this year's Musexpo on behalf of Cyanite after Covid...
How to Create Mood- and Contextual Playlists With Dynamic Keyword Search
In the last article on the blog, we covered how Cyanite’s Similarity Search can be used in music catalogs. In this article, we...
From Data to Decision – How to Use Music Data and Analytics for Intelligent Decision Making
We continue writing about the Data Pyramid and in this article we finalize the series with an overview of the fourth level of the...
Free Data Sheet: Full Tagging Analysis of Spotify’s New Music Friday
This analysis example gives you a comprehensive overview of Cyanite’s auto-tagging scope. Whether you are thinking about integrating...
Benchmarking in the Music Industry – Knowledge Layer of the Data Pyramid
This article is a continuation of the series on the Data Pyramid which allows to turn music data into actionable insights. In this...
Making Sense of Music Data – Data Visualizations
This article is a continuation of the series “How to Turn Music Data into Actionable Insights”. We will explore the second step of...