Every learner knows the cycle: you copy twenty words into a notebook, review them for three days, and a month later the notebook reads like a stranger’s shopping list. The words did not leave because your memory is weak. They left because they never had anything to hang on to — no voice, no scene, no moment, just a row in a list.
Words met inside a real video behave differently. You heard them at a specific second, in a sentence that was going somewhere, spoken by a person who meant it. This guide is about turning that encounter into permanent vocabulary: which words to capture, how to bind them to their moment, and how to review so they survive.
A word needs a moment, not a row
A dictionary entry gives a word one cue: its written shape. A word in a video arrives with four or five at once — the sound, the subtitle line, the scene on screen, and the story around it. From a breakfast-diary vlog in the corpus:
サッと食べるね。やっぱり朝子供たちもバタバタだし
“You can eat it quickly — mornings with the kids are hectic, too.”
When you later try to recall 食べる, you are not recalling a row. You are recalling a mother eating fast because the morning was chaotic — and that memory carries the casual ね ending and the し-chained reason along with it, free. One good sentence can teach a verb, a register, and a grammar pattern in a single pass. No list row does that.
The capture loop
The method is a loop, not a session. Each step takes seconds, and together they move a word from “met” to “owned”:
- Meet the word in context. You are watching a video you mostly understand; one word blurs past. That word — not a pre-made list — is your vocabulary for the day.
- Capture it at its moment. Tap the word while the subtitle line is still on screen. The word, the full sentence, its translation, and the exact audio timestamp are saved together. In LingoDew this is one tap, and the player keeps playing.
- Let the scheduler do the remembering. The saved word enters FSRS spaced repetition and comes back tomorrow, then in three days, then next week — each time just before you would have forgotten it. The comparison of FSRS with the classic SM-2 scheduler explains why this timing matters: recall at the edge of forgetting is what strengthens memory most.
- Meet it again in the wild. Because the word was worth capturing, it will reappear in your next videos. The second and third encounters cost no study time at all — they are the review.
The loop’s power is in steps 2 and 4 pairing. The review card plays the original audio moment, so every review rehearses the sound, and every new wild encounter confirms the card. Vocabulary stops being a separate activity from watching.
Which words deserve the loop
Not every word is worth a card, and the corpus can tell you which ones pay back the effort. Across the 131 Japanese videos currently published, 食べる (to eat) is spoken 769 separate times in 88 videos — roughly two videos in three. 飲む (to drink) appears in 39 videos, ご飯 (rice, a meal) in 42, 買う (to buy) in 63, and 起きる (to wake up) in 24.
That ranking is study advice:
| Word | Heard in | Verdict |
|---|---|---|
| 食べる | 88 videos, 769 times | Active loop — capture every inflection you meet |
| 買う | 63 videos | Active loop — the errand verb |
| ご飯 / 飲む | 42 / 39 videos | Capture once, then let wild encounters do the rest |
| 起きる | 24 videos | One good chunk is enough for passive recognition |
Words heard in dozens of videos deserve the full loop, because every card you review will be reinforced again for free within a week of normal watching. Words heard a handful of times are usually not worth a card at all — look them up, move on, and if they truly matter, the corpus will bring them back to you.
Each word page on this site keeps the live counts and lists the real sentences behind them, so you can check a word’s weight before you decide to carry it. The N5 daily-life words guide walks through five of these words line by line.
Review with the sound attached
When the review queue resurfaces 食べる, the worst card shows you 食べる → “to eat”. The best card plays half a second of a real voice saying サッと食べるね and asks your ear for the meaning. You will meet this word as sound in videos for the rest of your life, never as a flashcard — so review it as sound.
That is also why the capture step saves the sentence and not just the headword. A line from the corpus:
まず本屋で教科書一冊買いました。
“First I bought one textbook at a bookstore.”
One card for 買う that carries this sentence also carries 本屋で (where), 一冊 (how the counter works), and the polite-past report shape that errand talk always takes. Reviewing the word in its sentence rehearses all of it at once, which is why sentence-bound cards wear better than isolated ones over months.
When a word is finally yours
A word is not yours when you can recite its meaning. It is yours when you hear it mid-sentence, in a new voice, at full speed, and your brain does not flag it as vocabulary at all — it is just part of what the person said. The loop gets you there because every stage rehearses the real form: captured in a sentence, reviewed with its audio, re-met in new videos.
Start with a genre where the same words circle back daily — routine and food vlogs. The N5 daily-life collection and the N5 food collection are stocked with exactly that vocabulary, and a video like the fifteen-minute breakfast diary will offer up 食べる, 飲む, and ご飯 before the second meal. Watch, tap, capture — and let the loop keep what the video taught you.