Detect Music Genre, Mood & Instruments Online

For broadly different genres (classical vs metal vs hip-hop) the top-1 match is right 85-95% of the time. For closely related styles (jazz vs blues, house vs techno) the model is right about 55-70% of the time, but the correct answer is in the top-3 about 80% of the time. The tool intentionally shows top matches with scores rather than a single verdict.

These two genres overlap musically (12-bar progression, blue notes, swing feel), and even experienced listeners disagree on borderline tracks. The model usually gets it right for clearly identifiable jazz or blues, but on jazz-blues fusion the top-3 list will list both and the percentages will be close. That gap itself is useful information.

Yes. The underlying model recognises musical structure, not language. K-pop, J-pop, Russian rap, Arabic pop, Latin music, world traditional pieces all produce sensible genre, mood and instrument outputs.

The music model is downloaded on your first click ("Analyze track"). It is then cached by your browser for six months, so subsequent analyses on the same device start instantly.

LAION CLAP (Contrastive Language-Audio Pretraining), the largest open-source music classifier available under the Apache 2.0 license. The text-side embeddings for every label are precomputed at build time, so your browser only loads the audio encoder.

After the first analysis the model is cached by the Service Worker and the page works offline. Subsequent analyses run entirely on your device.