How to Learn a Song by Ear, Step by Step

Set up before you start guessing

Three things, and they take under a minute between them.

1. Tune. Transcribing on an out-of-tune instrument means comparing your wrong note to the record and concluding the record is strange. Twenty seconds: tune the guitar, or whatever you play.
2. Check the recording is at concert pitch. If your in-tune instrument sounds consistently a little off against the track, the track is probably not at A = 440, common with older records. Shift the recording rather than retuning yourself; see the reference-pitch guide.
3. Find the key. This is the single highest-value minute in the whole process, because it takes the twelve possible notes at every decision point and makes seven of them likely and five unlikely. The method is in finding the key of a song by ear.

Take the bass line first

This is the step that turns the job from hard into tractable, and it is the one almost nobody does first.

The bass is the easiest line in any recording to hear. It is usually the loudest thing in its frequency range, it moves slowly, and it has very little competing with it. And it is enormously informative: the bass note under a chord is nearly always that chord's root. So an afternoon's work identifying twelve possible chords becomes a quick pass identifying six or seven bass notes, each of which then narrows the chord above it to two or three candidates.

Practically: loop four or eight bars, drop the speed to somewhere around 60 percent, and hunt for each bass note on your instrument one at a time. Sing it first if you can: singing a pitch before finding it is far more accurate than hunting blind, because it forces you to commit to what you actually heard.

If the bass is buried under everything else, cut the highs. Rolling off treble on an equaliser leaves the low end exposed and can make an inaudible bass line obvious. This is the same technique used to build a practice backing track, applied to listening rather than playing.

Then the chords, as a two-way test

With the roots in hand, you are no longer identifying chords: you are testing them, which is a much smaller task.

1. Play a major chord on the root, against the recording. Then a minor. One will sit inside the track and disappear; the other will grate.
2. If neither is right, add a note. The usual suspects are a dominant seventh (very common in blues, rock and anything with a turnaround), a sus4 (a chord that sounds unresolved and wants to move), or a minor seventh (jazz, soul, most things with a groove).
3. Sanity-check against the key. If your third chord requires a note from outside the key, it might be right and interesting, or you might have the wrong root. Check the bass note again before you commit to an exotic chord.
4. Write the progression down as you go. Roman numerals or letters, whatever you read faster. Progressions repeat, and once you have written out a verse you will often find the chorus is the same four chords in a different order.

Now the melody or the solo, two bars at a time

Everything above was setup. This is where the loop and the speed control do the actual work.

1. Loop a small section: two bars, four at most. A tight loop repeating endlessly is the tool that makes this possible, because it lets you hear the same phrase fifteen times without touching anything. There are three ways to do this in a browser in looping part of a video.
2. Slow it to where notes become separate events. Usually 50 to 75 percent. Keep the pitch locked: a slowed video that also drops in pitch is teaching you the wrong notes. The how is in slowing a video down.
3. Get the first note before anything else. Just one. Sing it, find it, write it. The first note anchors everything after it, and most of the difficulty in a phrase evaporates once you know where it starts.
4. Then work in intervals, not absolute pitches. Is the next note higher or lower? By roughly how much: a step, a skip, a leap? Your ear is far better at judging distance between two notes than at naming either one, and this is the skill that gets dramatically faster with practice.
5. Add the rhythm separately. Once you have the pitches, tap or count the rhythm against a click. Trying to get pitch and rhythm simultaneously is the thing that makes people give up. Two passes, one job each.
6. Speed the loop back up in stages and check your version still fits. A phrase that is right at 60 percent and wrong at full speed usually has a rhythm error, not a pitch error.

Check yourself at the end

Two checks, both cheap, both skipped by nearly everyone.

Record yourself playing your version over the track. Playing along and listening critically at the same time is not really possible: your attention is spent on execution. On playback, wrong notes and rhythmic misreadings that were invisible while you were playing become obvious immediately. This is the whole argument in recording yourself, and it applies to transcription more sharply than to anything else.

Then look at the tab. Afterwards, not during. The point is not to be told the answer, it is to find out where your ear went wrong, and the pattern in your mistakes is the useful output. If you consistently hear minor sixths as fifths, that is a specific thing you can practice. If you were right about everything except the rhythm, that is a different specific thing. Neither of those insights exists if you read the tab first.

Why it gets so much faster

Because most music is made of a small number of recurring shapes. There are perhaps a dozen chord progressions that account for a large fraction of popular music, a handful of common melodic contours, and a set of rhythmic figures that turn up everywhere. The first time you transcribe a I–V–vi–IV progression you work it out chord by chord over ten minutes. The twentieth time you recognise it in two bars and spend your effort on what makes this particular song different.

That is why the honest advice is quantity over difficulty. Ten simple songs will build the ear faster than one hard one, and the tenth will take a fifth of the time the first did. If it is still slow, you have not plateaued: you have not done enough of them yet.

Start with something you already know

Not something you can play, something you can sing from memory. Familiarity does an enormous amount of the work, because your ear already knows where the song is going and is only being asked to name what it expects. A nursery rhyme, a riff you have hummed for years, the bass line from a song you have heard four hundred times.

Do that one this week. Then do nine more. The skill is not mysterious and it is not a talent: it is a small method plus repetitions, and the repetitions are the part everybody skips.