How to shadow: a step-by-step method for your first session

Most advice on shadowing stops at “listen and repeat at the same time”, which is true and useless. What follows is an actual routine: what to load, how to cut it, how many times to say each piece, and how to know when to move on.

If you have not met the technique before, start with what shadowing is — this guide assumes you know why the audio does not stop.

Step 1 — Pick material you almost understand

The single most common mistake is choosing audio that is too hard. The target is material where you follow the meaning without a transcript but could not produce the sentences yourself. If you are reaching for a dictionary every few seconds, the recording is teaching you vocabulary, not speech.

Length matters as much as difficulty. Sixty to ninety seconds of audio is a full session. That sounds absurdly short until you have repeated it thirty times, at which point it is plenty. A five-minute clip guarantees you will practise the first minute and abandon the rest.

One voice is easier than several. Interviews with crosstalk, laughter and overlapping speakers are poor material no matter how interesting they are.

Step 2 — Cut it into phrases, not sentences

A phrase is what a speaker says between two natural breaths — usually two to six seconds. That is the unit you can hold in working memory and repeat as a single gesture. Whole sentences are frequently too long; by the time you reach the end you have lost the intonation of the start.

Cut at the pauses the speaker already left. They are not arbitrary: speakers breathe at grammatical boundaries, so the pauses tend to fall exactly where the meaningful chunks end.

In Shadowly this is what the waveform is for — the pauses are visible as flat stretches, and you drop a boundary in the middle of each one. On a subscription, auto-split finds them for you and places every boundary at once; the threshold is derived from the recording’s own speech level, so a quiet podcast and a loud one both split sensibly.

Step 3 — Listen once without speaking

One pass through the whole clip, no talking. You are checking that you understand it and letting your ear register the melody. Skipping this means the first several repetitions are spent working out what was said, which is not what the exercise is for.

Step 4 — Work one phrase at a time

Set the phrase to loop with a gap after it. A gap roughly as long as the phrase is the right starting point — long enough to say it back, short enough that you cannot dawdle.

Then take the phrase through three passes:

1. Read-along, if you have a transcript. Speak with the audio while looking at the words. This is the easiest version and it exists to get the phrase into your mouth at all.
2. Shadow it blind. No text. Speak over the audio, a beat behind. Expect to lose the thread in the middle the first few times; keep going rather than stopping to fix it, because recovering mid-phrase is itself the skill.
3. Say it into the gap, alone. Now the audio has finished and you produce the phrase from memory, matching the rhythm you just heard. This is the hardest of the three and the one that transfers to speaking.

Five to eight repetitions per phrase is usually where the returns flatten out. Twenty is not better; it is where attention goes and you start producing the sound without listening to it.

Step 5 — Record yourself and actually listen back

This is the step people skip, and skipping it is why learners plateau with a fossilised accent. You cannot hear your own voice accurately while you are producing it. Your skull conducts sound to your ears directly, so what you hear in your head is not what anyone else hears.

Record one phrase, then play the original and your version back to back. Do not listen for whether you got the words right — you know that. Listen for three things specifically:

• Stress. Is the emphasis on the same syllable, and the same word? Misplaced stress is more damaging to intelligibility than a wrong vowel.
• Rhythm. Did you space the words evenly where the speaker crushed some together? Learners tend to give every word equal weight; native speech does not.
• Pitch at the end. Rising and falling contours carry meaning, and copying them wrong makes a statement sound like a question or a polite remark sound curt.

Two or three comparisons per session is enough. The point is calibration — teaching your ear what your mouth is actually doing — not producing a perfect take.

If you do only one thing differently after reading this, make it this step. Repetition without playback is practice with the feedback removed.

Step 6 — Run the whole clip

Once the individual phrases are comfortable, shadow the clip end to end without stopping. This is the assembly step: phrases that were fine in isolation often collide at the joins, where one runs into the next without a breath.

Two or three clean full passes and the session is done.

How long, and how often

Fifteen to twenty minutes daily beats two hours on Sunday. This is motor learning, and motor learning consolidates across sleep; frequency does more than duration. A session that leaves you slightly tired around the mouth and jaw is the right intensity.

Stay on one clip for several days rather than chasing new material. The instinct to move on early is strong and wrong — the gains arrive on repetitions four through ten of a clip you already know, not on the first pass through something new. A clip is finished when you can shadow it end to end at full speed without losing the thread, which usually takes three to five sessions.

Practising away from the screen

The commute is good shadowing time, and it is where a prepared file earns its keep. Shadowly can bake the whole session — every phrase, with its repeats, its gaps and any speed change — into a single MP3 you can put on a phone. The practice then needs no interface at all: press play and talk.

Common mistakes

• Reading the transcript the whole time. It turns the exercise into read-aloud practice. Text is a crutch for pass one only.
• Speaking too quietly. Mumbling lets you skip the sounds you find hard. Speak at a volume you would use to talk to someone across a room.
• Slowing the audio down permanently. Half speed is a reasonable temporary aid on a genuinely difficult phrase, but the time pressure is the training stimulus — remove it and you are doing a different, easier exercise.
• Chasing perfect accuracy on phrase one. Get through the clip at a rough approximation, then tighten. Perfectionism early means you never reach the end.

Start

Take a ninety-second clip of something you nearly understand and run it through the six steps above. The Shadowly editor handles the mechanical part — the splitting, the looping, the gaps, the recording — in the browser, with no account and no upload.

If you are not sure what to practise with, choosing a podcast episode that works covers how to find material at the right level.