As rightsholders continually find new ways to identify music online, uploaders keep finding new ways to transform copyrighted content and evade detection. That’s why it’s important to adapt technology to solve ever-evolving issues, and why we’ve just released our enhanced cover song identification technology. To ensure we find the most uses of compositions, even if melodies, lyrics, or languages change, we’ve enhanced our cover song identification with phonetic matching and increased our true positive rate for melody matching by 35 percent. This allows us to more accurately identify cover songs, including compositions that may be used in AI-generated content leveraging voice clones, vocal swaps, or genre changes. With these enhancements, we’ve ensured our technology is the best on the market. Publishers, songwriters, and collection agencies can leverage this technology via our Search solution today. Dive in below to learn how enhanced cover song identification works.
Legacy cover song identification
By matching melodies between two or more recordings, we’ve been able to identify the use of compositions in digital content. For music publishers, this means finding cover versions and recordings of live performances.
Melody matching capabilities:
- Transpositions (key changes)
- Live concerts
- Short samples
- Different languages
- Different timbres and instruments
- Tempo changes or background noise
However, some cover versions may not use the original melody and instead only leverage a composition’s lyrics. In these instances, melody matching technology does not identify the covers or performances. In order to match lyrics, regardless of melody, genre, or language, additional technology is needed.
Enhanced cover song identification
In order to match lyrics accurately, we use digital fingerprints, which analyze the actual content in audio files, as opposed to analyzing transcribed lyrics/text. Digital fingerprints allow us to analyze the vocal expression of lyrics in a song, and match it against the vocal expression patterns in other audio files. We call this technology phonetic matching, since it’s based on phonetics and pronunciation, instead of written text.
Phonetic matching capabilities:
- Relies on vocal expression of lyrics, not extracted text or lyrics metadata
- Enables identifying songs with the same lyrics with very high accuracy
- Matches the same lyrics with different voices, genres, pitches, and tempos
- Language-independent and can identify compositions across different languages
For the most comprehensive solution, we combine our melody and phonetic matching technologies to provide enhanced cover song identification. With this approach, we can match lyrics despite different voices, genres, pitches, or tempos. By matching the vocal expression of lyrics, we can find cover songs that audio and melody matching alone would miss.
How enhanced cover song identification gets songwriters paid
Without advanced technology, heavily modified uses of compositions cannot be identified, so the royalties for these cover versions sit unclaimed in black boxes. But, with enhanced cover song identification, more of these missing royalties are unlocked.
- First, distinctive fingerprints are created by analyzing the phonetic and speech components of a recorded composition. Separately, fingerprints of the composition’s melody are created so both types of matching can occur.
- Next, these fingerprints are matched against our database of 120M+ recordings to find any matches containing the same phonetic or melodic qualities.
- Melody and phonetic matching work together to find uses of compositions that would be missed if using just one type of matching, or just audio identification.
- Every new match represents a previously unidentified use of the composition that can now be properly attributed and monetized by songwriters and publishers.
The future of music identification
As uploaders continue to find new ways to manipulate copyrighted music, including with AI generators, enhanced cover song identification is a powerful tool for music publishers and songwriters. By combining melody and phonetic matching, this technology ensures the most uses of compositions can be identified and the proper rightsholders are compensated for their works. We are constantly evolving our technology to meet the ever-changing challenges of digital content creation. Reach out to our team to learn more about enhanced cover song identification, or about how we’re safeguarding copyrighted music against AI.
