Shadowing with YouTube, without fighting the scrub bar

What shadowing actually is

You hear a sentence and repeat it aloud immediately, matching the speaker’s rhythm and intonation as closely as you can. Not translating it, not analysing it — reproducing it.

It came out of interpreter training, and language learners adopted it because it exercises several things at once that are usually practised separately: listening closely enough to parse, producing the sounds, and getting the prosody right. You cannot shadow a sentence you did not understand, which makes it self-checking in a way passive listening is not.

The pause is the whole exercise

The most common mistake is making the gap too short. People match the pause to the length of the audio, which feels logical and is wrong: producing a sentence in a language you are learning takes considerably longer than hearing one, especially early on.

A pause that cuts you off mid-attempt trains you to rush and to skip the sounds you find hard. A pause that is comfortably longer than the sentence lets you finish, notice what went wrong, and try again — which is where the learning is.

As a rule of thumb, give yourself about one and a half times the length of the line, and do not feel bad about needing it.

Why doing it manually fails

The technique is simple. The mechanics are what kill it.

By hand, a session is: play, listen, hit space, repeat aloud, drag the scrub bar back to a point you are guessing at, overshoot, drag again, play, listen. Every step is friction, and the dragging is the worst, because the scrub bar has no idea where a sentence starts.

Ten minutes of that and most people conclude shadowing is not for them. It is. The interface was.

Running a session

The subtitle box keeps the sentence and its translation on screen throughout, so you are never repeating something you do not understand.

1. Pick a video with clear audio and one speaker. Interviews and explainers are ideal; panel shows are not.
2. Choose a two to three minute stretch rather than a whole video. Shadowing is dense work.
3. Press Ctrl+H to start whole-video shadowing. The sentence plays, the video stops itself, and a countdown gives you the pause.
4. Repeat aloud. Actually aloud — subvocalising skips the part that is hard.
5. Let it continue by itself. If you finish early, press Skip rather than waiting out the countdown.
6. For a line that keeps defeating you, press H instead: it drills that single sentence, letting you hear it, repeat it, then hear it again.

When to add pronunciation feedback

Once the rhythm is comfortable, Say It Back adds a check: it listens, writes out what it hears in real time, and shows which words did not come through, in brackets inside the sentence itself.

Treat a low score as information rather than a verdict. Speech recognition is measurably worse on accented speech — which is what a learner produces — and on proper nouns. The word-by-word breakdown exists precisely so you can tell a genuine mispronunciation from the recogniser mishearing you.

It keeps the scores and the missed words so a pattern becomes visible over time. It never records your voice.

How much, and how often

Ten minutes of shadowing is more tiring than an hour of watching, and that is the point. Little and often beats a long session you never repeat.

A reasonable shape: two or three minutes of new material, then re-shadow a stretch you did yesterday. The second pass is where you notice how much easier it has become, and that feedback is what keeps people going.