Introducing Music Tagger V2

Music consumption has never been higher, mainly thanks to streaming platforms, from the big giants to niche platforms serving specific markets. As a result, the range of music genres consumed has greatly expanded, and it shows no signs of stopping. Local and even forgotten genres are now available for streaming, gaining new recognition — much to the listeners’ pleasure as well as sync opportunities.
In the same way, it’s not all about electronic and hip-hop hits; music supervisors need to tap into more classical and niche catalogs. A delicate harp chord might be the missing puzzle piece for a client’s project. An instrument detection technology must reflect all this acoustic authenticity and vibrancy, putting it in front of the widest audience.

What route have we taken?

We really wanted to boost our metadata tagging engine capabilities in order to reflect all this diversity by increasing the granularity of music classification. Extending the range of subgenres and instruments, as well as tracking their presence, was our priority. So, we put our MIR (Music Information Retrieval) algorithm back into training for these dedicated tasks.

As we strive for ethical artificial intelligence practices, we commit to training our classification model on publicly available or licensed tracks only. For this purpose, we used an open-source library to gather the most extensive list of musical genres and instruments to add to our directory. We then applied this data to our own metadata-dedicated musical catalog, which serves as the most robust and accurate ground of truth for music classification. By training our identification algorithm on this expertly labeled dataset, we’ve been able to deeply specialize in genre detection and score instruments' presence.

Ready for you to leverage!

Improvements overview

• 139 subgenres (from 12 main genres): Dive into subgenres for the most distinct flavor; looking for Soca more than Merengue or Mambo? We got you covered.
• 18 instruments: We now map the full string and wind spectrum, in addition to the drum types, bass, and orchestration. The mystical and airy character of the harp is just one click away.
• Comprehensive instrument presence data: The devil's in the details, and we've gone all in with statistics—providing overall, average, and median scores for each instrument, backed by standard deviation.
• Increased versatility: Capable of isolating sung vocals of all types, and even whispers in the background, from all genres of music, old and new, from classical pieces to industrial noise tracks.

Take the opportunity to test it right away

👉 Try it on your own files through the “Tasks” feature on your user dashboard