Chord detector

Chord detection listens to an audio file and returns the chords and progression section by section. No music theory or manual note-hunting on the fretboard required. Upload a track, the AI analyses pitch and harmony, and you get a chord chart ready to play along with. The tool is still being built and is not open to users yet.

What is automatic chord detection?

Automatic chord detection is a machine listening to a recording and estimating which chord the instruments are playing at each moment. It happens in three stages. First the track is cut into short frames and the energy at each pitch class is measured, producing a chromagram, a map of the twelve notes of the octave over time. Then that energy shape is matched against known chord templates to pick the closest candidate. Finally the sequence is smoothed over time, because real chords last for bars rather than jumping every instant. What comes out is a chord chart tied to timestamps, enough to play along with.

How accurate is automatic chord detection?

Fine for basic major/minor chords in a simple arrangement. Extended chords like 7ths, 9ths and sus chords, or the dissonant intervals common in jazz and modern pop, get misread as the nearest chord the model actually knows fairly often. The more layered the mix (drums, bass, strings, a dense vocal), the higher the error rate, since the model has to isolate harmony from everything else by ear alone.

Why separate the track before detecting chords?

A thick, heavily processed vocal often masks the frequencies of the backing instruments, which is exactly where chord detection tends to get it wrong, usually right at the chorus. Split the vocal from the instrumental first, then run detection on the instrumental alone; the result is noticeably cleaner than feeding in the full mix with vocals still in it.

What genres does this work best on?

Best on pop, ballads and acoustic tracks with clear harmony and few dissonant intervals. For jazz-level harmonic complexity or dense electronic productions, treat the output as a draft to check by ear, not a final answer.

How does this compare to working it out by ear or looking up a chord sheet?

The three approaches complement each other rather than replace each other. A published chord sheet is the most accurate option when a song is popular enough for someone to have transcribed it, but many sheets are user-submitted and still wrong, and new or obscure songs simply do not have one. Working it out by ear gives the most trustworthy result and trains your hearing, at the cost of time and some theory background. Machine detection is the fastest and works on any file including your own recordings, but it belongs in the draft category: the sensible use is to let it map out the harmonic skeleton, then sit at the instrument and verify the handful of transitions that matter, rather than trusting it wholesale or ignoring it.

Who needs automatic chord detection?

• Self-taught guitar or piano players who want to accompany a favourite song
• Musicians who need a quick chord reference for a song with no sheet available
• Music teachers preparing practice material for students
• Cover artists who need the original progression before rearranging it
• Anyone double-checking chords they picked out by ear

The tool is not open yet. These are the steps it is being built to have.

1. 1

   Upload the song

   Pick the track you want chords for. A full original recording gives better results than a vocals-only file.

2. 2

   Wait for the analysis

   The system estimates the key, the beat grid and the chord per bar, then lays them out along a timeline.

3. 3

   Read the chord chart

   Each chord is pinned to a timestamp in the song. Click a chord to jump to that moment and check it by ear.

4. 4

   Transpose to an easier key

   If the original key is full of accidentals, transpose the whole chart to something easier to fret and use a capo to compensate.

• The output is a machine estimate, not an authoritative transcription. Dense harmony and passing notes make neighbouring chords easy to confuse.
• Songs that modulate mid-way or use pentatonic scales are where errors cluster. Always check by ear.