How to Find the Chords of a Song (By Ear, by App, or with AI)

Every chord has a root, and in most popular music the root is the lowest note you hear in that bar. So the practical method is to listen for the bass, name that note, and then decide whether the chord above it is major, minor, or something with an extra note in it. Do that bar by bar and you have a chord chart.

Three routes get you there, and they are worth doing in this order. Work the chords out by ear, because that is the only version you can trust completely. Check the result against an existing chart if one exists, and treat the chart as a second opinion rather than as the answer. Use automatic chord recognition when the song is too dense, too new, or too obscure to have a chart and your ear keeps sliding off the harmony.

“The bass player is already telling you the root of every chord. Most of the work is learning to listen to the bottom of the mix instead of the top.”

These two questions get tangled together constantly, and separating them makes both easier. The key is the shortlist: seven notes, and the seven chords built on them, that a song draws from. The chords are the specific order in which the song visits that shortlist. If you know the key, the chords are a small set of candidates and you only have to work out the sequence.

Take G major. Its chords are G, Am, Bm, C, D, Em, and F# diminished. A song in G will spend almost all of its time on the first six, and most pop songs use only four of them: G, D, Em, C, in some order. Knowing that narrows the listening task from twelve possible roots down to a handful, which is the difference between guessing and checking.

The same seven notes also form E minor, the relative minor, with the same chords but a different home. That ambiguity is the single most common reason a transcription sounds almost right. It is worth reading a guide to finding the key of a songbefore you go further, because the key decision comes first and everything else hangs off it.

Find the tonal centre before you name a single chord. The fastest reliable test is the ending: play the last chord of the song and the bass note underneath it. Songs overwhelmingly resolve home, so the final chord is usually the tonic. If it feels unfinished rather than settled, try the relative minor three semitones below.

• Detect it from the audio with the BPM and key finder, then confirm the result by playing the tonic chord against the recording.
• Cross-check with a second reader. The audio key finder analyses the harmonic content independently; if two detectors disagree, the song is probably modal or modulating.
• Listen for the recurring chord. Whatever chord the song keeps returning to, especially at the end of a phrase, is the strongest candidate for the tonic.
• Hum the resolution. Sing what sounds like the "home" note over the outro. That note is the tonic, and its name gives you the key.

While you are at it, get the tempo too. It is not required for the chords, but a click at the right BPM makes bar-by-bar work much faster, and it stops you from losing your place in songs with long intros. TheBPM detector or thebeat counter will give you a number you can tap along to.

The bass is the most reliable guide to chord roots because it is usually playing one note at a time, it sits low enough to be separated from the vocal, and it almost always lands on the root rather than an inversion. Set the recording to loop one bar and listen only to the bottom end. Ignore the guitar, ignore the vocal, and try to hear a single pitch.

Then find that pitch on your instrument. Hum the note first, then match the hum. Going straight from recording to fretboard is where beginners lose time; the hum in between is the step that makes the match possible. Once you have the note, check it against the candidate chords in the key you identified. If the bass note is D and the key is G, the odds are strongly on D major or Bm depending on the chord quality above.

• Slowed playback helps. Drop to 75% or 80% speed without changing the pitch and the bass line stops blurring. The pitch changer and most browsers already keep pitch locked while the tempo moves.
• Separate the low end if you can. Running the track through a stem splitter gives you an isolated bass part, which turns a hard listening task into an easy one.
• Watch for inversions. When the bass note is not the root of the chord you expected, you are probably hearing the third or fifth in the bass. That is a slash chord, not a wrong note.
• Mark the bar, not the beat. Chords change on bar lines far more often than mid-bar. If a change sounds like it lands in the middle, check whether you are counting the pickup.

Work in short loops, four or eight bars at a time, and write down each root as you go rather than trying to hold the sequence in your head. A verse that feels impossible at full length is usually four bars repeated twice with one substitution.

You have a root note. Now determine the quality, which in most songs means choosing between major and minor. Play the major chord on that root, then the minor chord, against the recording. One will sound like it belongs and the other will sound like a disagreement. That test is crude and it works.

If neither fits cleanly, you are probably hearing a seventh, a suspended chord, or an inversion. Those sound ambiguous against a plain triad, and the fix is to try the likely extensions in order: dominant seventh first, then major seventh, then suspended fourth. In soul, jazz, and city pop the seventh is often the default rather than the exception, so start there.

“Major or minor is a question you ask the recording, not a question you answer from the melody. Play both, and let the track disagree with one of them.”

Write down what you hear, even when you are unsure. A chord chart with two question marks in it is worth more than a mental model, because the marks tell you where to listen again. If you are writing your own progression rather than transcribing one, achord generator is a faster way to explore what the same key can do before you commit to a sequence.

Once you have a root sequence, convert it to Roman numerals or Nashville numbers relative to the key. In G major, G becomes I, D becomes V, Em becomes vi, and C becomes IV. This single step makes everything after it easier.

• Patterns become visible. I, V, vi, IV is the most common sequence in pop. If your chart reduces to that shape, you have almost certainly transcribed it correctly.
• Transposing becomes trivial. The same numbers in any key produce the same progression, so a capo or a key change no longer means redoing the work.
• You can compare songs. Two tracks that look unrelated in chord names often turn out to be the same progression in different keys.
• The chart survives disagreement. If a smarter listener says the song is in B minor rather than D major, the relative numbers barely change.

Numbers also make it obvious when something is borrowed. A major chord on the flat-seven degree, or a major IV in a minor key, stands out from the pattern immediately, and those are exactly the moments worth checking twice by ear.

Crowd-sourced chord sites are useful and unreliable in the same breath. They are excellent for well-known songs, where dozens of people have corrected the same chart over years. They are close to worthless for new releases, deep album cuts, and anything without an active fanbase, because the only version available may be one person's first guess.

• Check the rating, not the listing. A version with many votes is usually right; a single unrated version is a hypothesis.
• Compare two versions. Where they disagree is exactly where the harmony is interesting, and where you should listen again.
• Expect a capo or transposed key. Charts are often written in an easier key to play, not the recorded key. Convert with numbers before you judge accuracy.
• Charts are not timing. Static text tells you which chords but not when they change. That is the part you still have to hear.

Treat every chart as a draft of your own transcription, not as a reference to be trusted. The version you build yourself, wrong in places, will teach you more about the song than a correct one you never checked.

Automatic chord recognition listens to audio and reports a chord sequence. The typical pipeline separates harmonic content from percussion, estimates how much energy sits on each of the twelve pitch classes in every beat, then compares those estimates against chord templates and picks the best match. The output is a chord chart aligned to time, which is the one thing static tabs cannot give you.

It is very good on clean, conventionally arranged recordings: plain triads, steady tempo, a bass that plays roots. It is noticeably weaker on jazz voicings, drop tunings, dense distortion, live recordings, and anything modal. The failures are predictable, which makes the tool useful as long as you know where to double-check.

• Roots are reliable, extensions are not. A ninth or a flat-five is usually reported as the plain triad.
• Slashes get flattened. Inversions often come back as the root-position chord.
• Boundaries drift. Changes are sometimes placed a beat early or late, especially on material that was not recorded to a click.
• Improvisation breaks it. Long solos over a static chord produce a lot of noise in the output.

The best use of detection is as a first pass over a song too long to transcribe manually. Let the tool produce a draft, then correct the passages you actually care about. If you want the harmony as editable notes rather than as chord names, the audio-to-MIDIroute gives you the pitch content and theMIDI editor lets you fix the misreadings directly.

Putting the steps in the right order saves an hour per song. This sequence works for almost any track.

• 1. Get the tempo and key. Two numbers that size the problem. Use the BPM and key finder and write both down.
• 2. Loop four bars at a time. Ignore the rest of the song until those four bars are settled.
• 3. Find the bass root in each bar. One note per bar, hummed before it is played.
• 4. Test major against minor. Two chords, one comparison, done.
• 5. Convert to numbers. Check whether the result reduces to a familiar progression.
• 6. Verify against a detection pass. Where the machine disagrees with you, listen a third time; usually one of you is right and it is worth knowing which.
• 7. Play it through. A chart that works bar by bar but fails end to end is usually missing a chord in the transition.

If you are working toward a release rather than a personal transcription, the same chord information feeds straight into the next stage. Write the progression into your own track, then openthe generator when you need the harmony rendered as audio rather than played, or thelyrics workspace when the melody needs words over the changes.

Some material defeats every method above, and it helps to recognise the reasons rather than assume you are doing something wrong.

• Modal songs. A track that keeps returning to D over a C major scale is in D Dorian, not C major. Chords alone will not settle it; the home note will.
• Constant modulation. A chorus that shifts up a whole step is common in pop and country. Chart each section separately and note the shift.
• Heavily distorted guitars. Power chords plus distortion produce strong overtones that read as extra pitch classes. Expect the detector to report sus and add9 shapes that are not there.
• Live and rubato recordings. No grid, drifting tempo, and changes that land off the beat. Loop smaller, and slow the playback down more than usual.
• Dense production. Layered synths and sidechained pads blur the boundary between chord and texture. A stem separation pass usually clarifies it.

“If a song resists three passes, the problem is usually that it does not have a single answer. Chart what you hear and say so.”

• Listening to the melody instead of the bass. The melody moves, the root does not. Follow the bottom of the mix.
• Assuming the first chord is the key. Intros often start on the IV or the vi. Check the ending instead.
• Skipping the hum. Going straight from recording to instrument nearly always produces a wrong first guess.
• Trusting one chart. Compare two sources and listen where they differ.
• Ignoring inversions. A bass note that clashes with your chord is information, not an error. It usually means the third or fifth is in the bass.
• Working at full length. Four bars at a time finishes faster than a whole song at once, every time.
• Forgetting to write it down. A transcription you cannot reproduce is not a transcription.

What is the fastest way to find the chords of a song?

If the song is well known, search for an existing chart first and check it by playing along. If it is new, obscure, or the chart sounds wrong, work it out from the bass line: find the key, then name the lowest note in each bar. Automatic chord detection is the fastest option overall but needs correction on anything harmonically unusual.

Can I find the chords of a song without any musical training?

Yes, with a keyboard or a guitar in front of you and a repeatable process. Find the key, listen for the bass note in each bar, then test major against minor. The skill you are actually building is hearing a single low pitch clearly, and that improves quickly with a few songs of practice.

Do I need the key before I can work out the chords?

It helps enormously and costs a minute. The key reduces the candidates from twelve roots to seven chords, and in practice to four or five. You can transcribe without it, but you will spend much longer deciding between options that the key would have eliminated.

How accurate is automatic chord recognition?

Very accurate on simple triads in clean, steady-tempo recordings, and noticeably weaker on jazz voicings, distortion, inversions, and live material. Treat the output as a draft: the roots are usually right, the extensions and the exact timing are the parts to check by ear.

Why does a chord chart sound wrong when I play it?

Three usual causes. The chart may be in a different key for easier playing, in which case convert it to numbers and transpose. It may be missing extensions such as sevenths or suspensions. Or the recording may use an inversion, where the bass note is not the root, which a plain chord chart cannot express.

Should I write chords as letters or as numbers?

Both. Letters are what you play; numbers are what you understand. Numbers reveal the progression pattern, make transposition trivial, and let you compare songs that look unrelated in chord names. Most working musicians keep the number version in their head and the letters on the page.

How do I find the chords of a song by ear alone?

Loop one bar, hum the lowest note you can hear, then match that note on your instrument. Play a major chord on it, then a minor chord, and keep whichever one the recording agrees with. Move to the next bar and repeat. Once you have the sequence, convert it to numbers and check whether it forms a pattern you recognise.

Is transcribing chords from a recording legal?

Working out chords for your own study, practice, or performance is normal musical activity and raises no issue. Publishing a transcription as sheet music or tab is a different matter, because it reproduces someone else's composition; check the publisher's terms before distributing one.