Drop any mp3 / wav / flac / ogg, any length: only the first 30 seconds are used. Or type a song name. Everything runs in your browser, nothing is uploaded.
Drop an audio file here or click to choose (first 30 s are analysed)
Loading the model (21 MB, once)…
A 5.5M-parameter CNN on log-mel spectrograms, trained on 26k Deezer previews for a Harvard first-year seminar. Eight genres: classical, pop, jazz, hip-hop, rock, electronic, country, film score. Test accuracy 77 %, 79 % on 500 clips from unseen artists. Code and report: github.com/Guczyslaw/fys-audio-ml.