Audio to chords
Any audio file your browser can play, EarChord can read: MP3, WAV, M4A, AIFF, FLAC, OGG. Drop one in and get the progression, the key and the tempo.
How the detection works
The audio is cut into overlapping windows of about 190 milliseconds. Each window is transformed into a spectrum, and every frequency component between 55 Hz and 2 kHz is folded onto the twelve pitch classes. The result is a chromagram: how much C, C sharp, D and so on is sounding at each moment.
Each frame of that chromagram is compared against 24 templates, one major and one minor triad per root, and the closest match wins. A short mode filter removes single-frame flickers, and neighbouring frames with the same label are merged into the segments you see on the timeline.
What it will and will not tell you
The vocabulary is deliberately 24 major and minor triads. Widening it to sevenths and suspensions is easy to do and measurably makes ordinary pop and rock worse, because the extra templates win frames they should not. A song full of sus4 chords will come back as the nearest triad.
Inversions are not reported either. A C major triad with E in the bass is labelled C, which is the chord most players want written on the page anyway.
Questions
- What accuracy should I expect?
- On clear recordings with one or two harmonic instruments, most chords land. On dense, loud, heavily produced mixes it degrades, and sections with no harmonic content at all are guesswork. The timeline shows what was detected so you can check it against your ear.
- Can it separate instruments first?
- No, and that is on purpose. Source separation means running a neural network, which is not something a browser tab can do for free. Everything here is signal processing that finishes in a fraction of a second.
- Do I need to sign up?
- No. No account, no email, no daily limit.