Shadowing is repeating Japanese audio out loud while it plays, a fraction of a second behind the speaker. Done for ten minutes a day with sentences at your level, it trains the rhythm, intonation and speed that grammar study does not reach. Below is a routine you can follow today, with practice audio.
In short
- Pick short audio you understand, with a transcript.
- Listen, read, mumble along, shadow, then record one take and compare it with the model.
- Research supports gains in fluency and intonation; evidence on individual sounds is mixed, so pair shadowing with feedback where you can.
What do you need before you start?
Three things: audio at your level with a Japanese transcript, earphones, and a way to record yourself (your phone’s voice recorder works). The audio matters most. If you cannot follow most of the transcript, pick something easier, or you will be copying sounds rather than practising Japanese you can use.
Good sources: textbook dialogues you have already studied (they come with audio and text), the practice sentences on this page, and short scenes you know well with Japanese subtitles.
The 10-minute shadowing routine
Listen without the text 1 minute
Play the sentence two or three times. Notice where the voice rises and falls and where it pauses.
Read the transcript 2 minutes
Check every word and the meaning. Look up anything you do not know; shadowing is not the time to guess.
Mumble along with the text 2 minutes
Speak quietly at the same time as the audio, reading as you go. Use the slow version if you keep falling behind.
Shadow without the text 3 minutes
Look away and repeat out loud, about half a second behind the voice. Copy the melody, not just the words. Repeat until you can keep up three times in a row.
Record one take and compare 2 minutes
Record yourself saying the sentence alone, then play the model and your take back to back. Pick one difference to fix tomorrow.
Practice sentences with audio
Six sentences, three at JLPT N5 level and three at N4. Each has a normal and a slow version. Start with the slow one if the normal speed feels fast.
Audio is generated with Google Cloud text-to-speech (a Japanese neural voice), the same kind of voice FluencyTool uses. It is not a recording of a person.
今日はいい天気ですね。
Kyō wa ii tenki desu ne.
Nice weather today, isn’t it?
The ね at the end rises a little, as an invitation to agree.
すみません、駅はどこですか。
Sumimasen, eki wa doko desu ka.
Excuse me, where is the station?
The u in desu is almost silent: “des ka”.
毎朝コーヒーを飲みます。
Maiasa kōhī o nomimasu.
I drink coffee every morning.
コーヒー has two long vowels: hold each for two beats (ko-o-hi-i).
雨が降ったら、家で映画を見ます。
Ame ga futtara, ie de eiga o mimasu.
If it rains, I’ll watch a movie at home.
The small っ in futtara is a one-beat pause before the t.
日本語を話すのは難しいけど、楽しいです。
Nihongo o hanasu no wa muzukashii kedo, tanoshii desu.
Speaking Japanese is hard, but it’s fun.
A long sentence: shadow the first half until it is smooth, then add the second.
もう少しゆっくり話してもらえませんか。
Mō sukoshi yukkuri hanashite moraemasen ka.
Could you speak a little more slowly?
Useful in real conversations. The ん in moraemasen is a full beat.
What should you listen for when you compare?
| Sound | What goes wrong | Example |
|---|---|---|
| Long vowels | Held for one beat instead of two | コーヒー (kōhī), not “kohi” |
| Small っ | Skipped, so the word loses a beat | ふったら (futtara), not “futara” |
| ん | Swallowed instead of held for a full beat | ませんか (masen ka) |
| Pitch | English-style stress on one syllable instead of a high and low pattern | あめ (rain) starts high and drops; see the pitch accent guide |
| Rhythm | Uneven syllable lengths; Japanese beats (mora) are close to equal | にほんご is four beats: ni-ho-n-go |
Common shadowing mistakes
- Audio that is too hard. If you cannot follow the transcript, you are not shadowing yet. Drop a level.
- Only mumbling. Quiet practice is a warm-up. Step 4 has to be at full voice.
- Never recording. You cannot hear your own speech accurately while you speak. The recording in step 5 is where you find what to fix.
- Too much material. Six sentences done well beat a five-minute podcast followed at half speed.
Does shadowing actually work?
The clearest evidence is for speaking fluency and prosody (rhythm and intonation). A 2025 systematic review of 44 studies by Whitworth and Rose found those gains most consistently, found listening gains mainly among lower-level learners, and found mixed evidence for the accuracy of individual sounds. Many studies used controlled tasks, and few tested whether the gains lasted. Our plain-English summary of the review covers what it found and its limits.
Shadowing is one part of speaking practice, not all of it. It does not replace conversation, where you have to produce your own sentences. The guide to practising Japanese speaking shows how shadowing fits with conversation and daily output, and the comparison of speaking practice apps covers tools for each part.
Frequently asked questions
What is shadowing in Japanese?
Shadowing means repeating Japanese audio out loud while you listen to it, with little or no delay, as if you were the speaker’s shadow. It trains rhythm, intonation and the speed of your speech, because you copy a model in real time instead of building sentences from grammar rules.
How long should I shadow each day?
Ten minutes of focused shadowing is enough to start, and short daily sessions are easier to keep up than long weekly ones. Spend most of the time on a few sentences you understand, not on a long recording you can only follow halfway.
Can beginners do shadowing?
Yes, if the audio matches your level. Use short sentences you can already read, slow them down if needed, and look at the transcript while you shadow. Shadowing audio you do not understand mostly trains mimicry of sounds, not Japanese you can use.
Should I look at the text while shadowing?
Use the text first, then drop it. Reading along helps you catch every sound in the first rounds; shadowing without the text afterwards makes you rely on listening, which is the skill you are training.
Does shadowing improve pronunciation?
Research mostly supports gains in speaking fluency and prosody (rhythm and intonation). Evidence for the accuracy of individual sounds is mixed. A 2025 systematic review of 44 studies by Whitworth and Rose found this pattern; our summary of it is linked below.