musiio
YEAR: 2020
WHAT IS Musiio?
Musiio uses Artificial Intelligence to help you automate your music library workflows. Tag, search and playlist large catalogues of music with ease using Musiio's ready-to-go solutions. This allows you to be even more creative when it comes to playlists and helps you serve the listener in a better way.
What makes Musiio so special? They apply AI to the audio file itself, which allows them to turn the audio file into mathematical features and train the brain of the tool. So when trying out a song, you receive information about the genre, mood, energy, instruments, emotions and so on.
TEST CASE with VRT Sandbox
What was our purpose?
At VRT, our main goal and purpose were to investigate whether we could enrich the VRT music database with “objective” data (not interpreted differently by different people) that are not available at this moment. We also wanted to see how we could use the genre-analysis to analyze the music of the different music channels and if the app offered added value to be used at VRT for specific target groups. Our last goal was to find out whether tagging could be useful to detect “doubles” in our music database (e.g. same song but with a different title or master).
How did we test?
A manual selected list of almost 17.000 tracks, with different genres, that were played on the five different VRT radio stations on Thursday the 25th of June 2020 were forwarded for tagging. Besides this batch, we decided to add another large amount of instrumental music accompanied by a limited selection of archive recordings for tagging.
What were the results?
The tracks were all tagged by Musiio and are searchable in the Musiio app. Besides that, Musiio also delivered an Excel document with all the tags and respective scores. These results could also be easily linked to the data in Zappa (our current Music library) by using the ID’s.
LEARNINGS
All the purposes were successfully tested and provided us with different valuable insights. Musiio allows us to extend our music library with objective data, including both data about vocal presence, vocal gender, moods, instruments and so on. However, at the time we did experience some wrong interpretations of the beat feel, but mostly the interpretations were accurate.
For statistical research, the data that Musiio delivers on genre, seems to be very useful for the pop channels. Musiio provides each song with 1 to 4 genres, each with its own score. This makes it possible to use the genre with the highest score in specific radio programmes/channels, or to make more subtle calculations.
The app could provide a great added value as a secondary database working specifically for sonography. However, as a search system for an archive (where more context is often needed) this interface is certainly not the best solution.
In the Musiio app you can search for “similar tracks”. This is done based on the feature extraction where the features are compared with each other. The similarity search was however disappointing and gave us many false positives, and evident relationships were much less established.
Go to the website of Musiio ->