Go back to blog


Next-gen MRT: how melody and phonetic matching identify more music

Jake Handy

Millions of dollars of mechanical royalties should be going to songwriters, but instead go unmatched and unpaid. Why? Because most music technology companies and royalty organizations are relying on only audio and metadata matching. 

Metadata matching is insufficient due to poor data quality, and matching becomes inconsistent when key identifiers are missing entirely, such as ISWC, ISRC, even song titles. Audio matching can typically match together songs more accurately than metadata, but it’s still not enough for songwriters that rely on data linking between audio recordings and their silent composition counterparts. And while audio matching can find identical recordings and accommodate modifications like pitch or speed changes, it’s not enough to identify cover songs, changes in genre or instrumentation, or significant changes in lyrics. 

What do all of these unidentifiable examples have in common? They are all uses of compositions, and they all prevent songwriters from getting paid. To identify these uses and pay out more mechanical royalties, additional Music Recognition Technology (MRT) is required. 

At Pex, we combine audio fingerprinting and matching with the power of melody and phonetic matching to identify content with profound modifications. This comprehensive approach identifies the most music, which enables unmatched royalties to be paid out accurately, and gives rightsholders greater control over uses of their content for monetization or anti-piracy. Take a look at some of the content we’ve matched below. 

Matching the same composition with different instrumentations 

Melody matching identifies content based on the underlying composition (repeating notes, melodies, chord progressions, etc.). Using our melody matching technology, we identified two different uses of the classical composition Nocturne No. 1 in B-Flat Minor. One version of the song is played on the piano, and another version is played on the violin. You can hear that this is the same composition performed on two different instruments with very different tonal qualities. If we were using only audio matching to detect identical recordings of the piano version, we would not have discovered the violin version. 

Listen to the piano recording by Chiharu Aizawa

Listen to the violin recording by Marte Krogh

Matching the same composition with lyrics in different languages

Also using our melody matching technology, we identified two uses of the same composition with lyrics recorded in different languages. This version of “How He Loves Us” by The Worship Initiative and Shane & Shane is sung in English, while this version of the same song is sung in Spanish. If the English composition or recording is what’s registered with royalty collection organizations, the Spanish version of this song will likely not be identified without the use of something like our melody fingerprints, and the composition owners will go unpaid for this use. 

Listen to the English version 

Listen to the Spanish version

Matching two different recordings with the same lyrics 

Phonetic matching identifies content based on speech or the vocal expression of lyrics (the same words being sung regardless of intonation or inflection). Using our phonetic matching, we identified this version of “Simply the Best” by Billianne and matched it to the original recording by Tina Turner. The original recording uses multiple guitars, keyboard, and backing vocals for a much more produced sound. Billianne’s version uses only acoustic guitar and her vocal style is drastically different from Turner’s. A cover version as different as Billianne’s wouldn’t be detected by audio or melody matching, but because the lyrics are the same in both versions, we identified it with phonetic matching. 

Listen to “Simply the Best” by Tina Turner

Listen to “Simply the Best” by Billianne

Matching the same lyrics with different vocal styles

Using only our phonetic matching, we were able to match this a cappella version of “Feeling Good” by Dominique Fils-Aimé to a similar recording by Michael Bublé, which uses instruments. The two recordings have very different vocal styles, and the Bublé version uses background music in places the Fils-Aimé version does not, but the two covers still matched thanks to their use of the same lyrics. 

Listen to “Feeling Good” by Dominique Fils-Aimé

Listen to “Feeling Good” by Michael Bublé

Match more with Pex

At Pex, we’re the leaders in digital rights technology. We use audio, video, melody, and phonetic fingerprints to identify music across the web, so rightsholders can be credited and compensated. If you’re relying on only audio or metadata matching, you’re missing many uses of your content. And if the organizations you use to collect royalties or claim your content for you are relying on just these two resources, they’re also missing uses. Stop letting your content and money go unmatched. Reach out to schedule a demo and see how you can start leveraging Pex technology today.


Recent stories

See more Technology