ElevenLabs has released Music v2.5 for its AI music generator. The version is available via app and API, offering users free and professional tiers.
According to The Decoder, in a blind test with nearly 48 000 comparison pairs, listeners more frequently chose the new version over the previous one. The company also states that the model was trained exclusively on licensed music.
The practical implication of the release is expanded access to music generation through various channels and pricing tiers. However, the source is presented as an independent retelling of metadata rather than a full publication describing the test, meaning the details of the comparison require separate verification.
editorial commentary
Why it matters
A likely consequence is increased accessibility of music generation for users and developers via the app and API. The next observable signal will be details on the testing methodology and independent evaluations of the new version. Significant uncertainty remains because the package contains only one retelling of metadata without a full description of the evidence.