Subscribe to immit jobs
Hand-picked jobs in Japan that need your Japanese. Written up in English, with visa and language details included. For N2+ learners.
How to Mine Japanese Vocabulary from Netflix and YouTube in 2026
You finish a Japanese show on Netflix or a YouTube interview in Japanese. There were a bunch of words you wanted to remember. A day later you can picture the scene but the words are already gone. The next video reuses two of them and you look them up again. This is the pause-rewind-forget loop, and most language learners I know have spent months inside it before finding a way out.
The way out has a name in the immersion community. It is called sentence mining, and most of this post is about how to actually run that loop in 2026 without spending a weekend on tool setup. The post also handles a fact I see learners learn the hard way: video mining surfaces are not all equal in 2026. YouTube is a clean fit for a popup dictionary plus SRS workflow. Netflix is not. We will cover both plainly, and the post my wife and I built immit for is the YouTube half.
How to mine Japanese vocabulary from YouTube in 2026
How to mine Japanese vocabulary from YouTube in 2026 comes down to four pieces: a Japanese YouTube video with real Japanese captions, a popup dictionary that turns any single word into a definition and reading, a one-click save to a sentence card, and a spaced repetition queue that brings the word back tomorrow. YouTube exposes its Japanese captions as selectable text on the page, which is what lets a popup dictionary read them. That single technical fact is why the YouTube workflow below works and why the Netflix workflow needs a different tool, which we will get to.
The rest of the post walks through the YouTube workflow end to end, the entities involved, and the trade-offs against Migaku, Language Reactor, asbplayer, and the classic Yomitan + Anki chain.
What is japanese sentence mining, and why it works for language learning
Sentence mining (sometimes called japanese sentence mining when the target language is Japanese) is the process of extracting sentences that contain unknown words from content you read, watch, or listen to in your target language, then saving each one as a targeted sentence card (often shortened to TSC) so the word and its context come back together at review time. The standard immersion workflow stores these as anki cards in Anki or as the equivalent inside an integrated tool. The point is not to memorize individual words in isolation. The goal of sentence mining is to find useful sentences and phrases that help solidify the meaning of a word as someone might use it in the real world, rather than studying words in isolation.
A useful framing for new learners: rather than studying isolated vocabulary lists, you find sentences in immersion content that are just barely out of reach, and let the surrounding context teach the new word for you. This is what language learners mean when they talk about mining sentences from native material. Over a few months, the habit reshapes your japanese learning from "study, then watch" into "watch, then save the words you actually want to keep."
A quick fact about why this works. Spaced repetition systems (SRS) use a schedule of expanding intervals to bring a card back for review just as the learner is about to forget it. immit's SRS is FSRS: each review updates the card's schedule from how well you remembered it, so the interval grows when you remember a card and shrinks when you forget it, which means you end up spending review time only on the words you are actually about to forget. When the schedule and the source material both come from your real life, the words you save are anchored to scenes the brain has already encoded, which is what makes the recall durable.
One thing to clarify up front. Sentence mining is a vocab and reading habit. It is not a substitute for speaking practice, and it does not replace structured lessons if your grammar foundation is shaky. In the past, learners often tried to use mining alone and stalled out on output skills; the modern approach is to use mining for vocab acquisition and to handle speaking and grammar lessons through other tools (italki tutors, Bunpro, Renshuu, or a textbook series). This post is about the vocab half.
Netflix vs YouTube as a mining surface in 2026
Netflix and YouTube look interchangeable to a learner. Both let you pause Japanese audio, both can show Japanese subtitles, both feel like obvious places to mine. The difference is technical, and it matters for which tool you can use.
YouTube renders its Japanese captions as selectable text inside the page. That means any popup dictionary that reads selectable text on a page will work on YouTube CC. immit's popup dictionary fits this pattern, and so do Yomitan and Language Reactor.
Netflix renders its subtitles inside a non-selectable player element. That means immit's popup dictionary does not currently work on Netflix subtitle text. Yomitan does not work directly on Netflix subtitles either, for the same reason. This is a real product limitation we want to state plainly rather than paper over. If Netflix is your only immersion source today, immit is not the right tool for that one surface in 2026. The serious paid alternative is Migaku, which ships its own native in-player overlay for Netflix specifically.
What this post does next is walk through the YouTube CC workflow in detail, because that is the surface where immit's lookup-and-SRS loop runs end to end. Section 9 of this post covers how a YouTube mining habit translates to the other Japanese surfaces where immit's hover lookup works, which is where the one-inbox design actually pays off over months.
The 2026 YouTube CC mining workflow, in five steps
Step 1: Find Japanese video with real Japanese captions
Open YouTube and pick a Japanese video that has Japanese captions enabled. Real Japanese captions, written by a human or by the channel, are different from auto-generated captions. The auto-generated layer often misses kanji, mangles loanwords, and breaks on regional accents, which makes the sentences less useful to mine. Look for the CC button on the video and check the language list. If "Japanese" is listed without "auto-generated," you have a clean source.
A practical step: create a dedicated YouTube account for Japanese content can improve your vocabulary learning by training the recommendation algorithm to surface more Japanese material. Watching only Japanese videos on one account compounds over weeks; the homepage starts pushing channels you have not seen, the suggested-next-video sidebar stays in Japanese, and the recommendation algorithm builds a feedback loop you do not have to maintain. This is one of the highest-yield half-hour decisions you can make for your japanese learning.
A second pacing note about the captions. Creator-submitted subtitles, which the channel uploaded themselves, are recommended over auto-generated captions for better accuracy on kanji and word spacing. Auto-generated captions also tend to drop punctuation and stumble on regional accents, which makes the sentences less useful to mine. The CC menu on each video shows whether the Japanese track is "auto-generated" or a real caption file from the creator.
Step 2: Install immit free in Chrome
Install immit free from the Chrome Web Store. The extension adds a popup dictionary that reads any selectable Japanese word on a page, including the YouTube caption pane. No account required. No setup beyond the install click.
Step 3: Pause on new words and hover the unknown word
Start the video. Pause on a line that contains an unknown word. Move your cursor over the Japanese word in the caption pane and the immit popup will show the reading, the definition, and example sentences. The lookup itself is near-instant. If the meaning makes sense in the context of the scene, you have a candidate to save.
Step 4: Save the sentence card in one click
In the immit popup, click save. The word, the reading, the definition, and the full Japanese sentence are stored as a single sentence card. There is no AnkiConnect step, no card template configuration, no deck management, no card template debate. The card lives in immit's spaced repetition queue and will resurface tomorrow for review.
Card template basics
The card template is the thing most Anki tutorials fight over and the reason many learners stall before they save their first card. immit's card template is fixed. The front of the card shows the target word. The back shows the definition, the pronunciation (reading), the part of speech, and the example sentence the word was saved from. There is nothing to configure, no template variants to choose between, no field mapping to set up. This is a deliberate design call: a fixed template removes the single biggest setup friction in Anki workflows, at the cost of the per-card customization Anki tinkerers prize.
If you ever want to migrate to a fully configurable system, the card content (the word, the reading, the example sentence) is portable; the card template you would otherwise have spent hours tuning is a sunk cost most people never recover anyway.
Find sentences worth saving (and skip the rest)
When you find sentences to save, choose ones where you know every word except one. To maximize retention, select sentences with only one unknown element when creating cards for spaced repetition, as this simplifies the learning process and enhances understanding of context. These are sometimes called i+1 sentences in the Refold framing. Short sentences with a single unknown element are easier to learn and easier to remember than long sentences with three unknown words and a grammar point you have not seen before. The goal is to find sentences where the surrounding context already makes sense and one new word is the only thing that needs to land.
Avoid picking machine-translated sentences when mining, and avoid sentences pulled from a context you do not personally find interesting. Using sentences from familiar sources greatly aids memorization, and if a word feels like filler or if you only half-care what the speaker said, do not save it. You will not save all the sentences worth saving, and that is fine; more sentences will come back. The mining habit lives or dies on whether the cards represent moments you cared about the first time through.
The sentence travels with the card
Every card you save from a video carries the full Japanese sentence it came from, so the context is there when the card comes back. The front of the card shows the word; the sentence and its source are on the card for when you need them. Cards mined from video and cards saved while reading the web live in the same immit review queue, and the dictionary definitions on the back use the same lookup that powered the popup. You do not have to guess at a meaning later because you saved the word in a hurry.
Step 5: Review the card the next day
Open immit the next day and the card will be waiting. The target word shows on the front; you recall the meaning and grade yourself, then the back of the card reveals the definition, reading, part of speech, and the sentence the word came from. FSRS will schedule the next appearance: the interval grows after each successful review and shrinks if you miss the card. Daily review of vocabulary through flashcard apps is what transitions a freshly saved word from short-term memory to long-term memory.
Building this into a study routine is what separates language learners who keep moving from language learners who stall. Five to ten minutes of review every morning is enough at the start. Active listening, where you focus on recognized or new words in context, makes the next watch session reinforce what you reviewed in the morning. Skipping review days is the most common reason mining stops working for people; it is more harmful than skipping immersion days.
The five steps above are the entire loop. There is no separate mining session, no batch import, no "I will set up Anki this weekend." You watch a video, you find sentences worth saving, you save four to eight cards in a sitting, and you review them later. Learning words this way, one episode or one video at a time, is what makes the habit stick.
What to do if Netflix is your main source
If your immersion content is mostly on Netflix and moving to YouTube is not realistic, there are two honest paths in 2026.
The paid path is Migaku. Migaku's Standard tier sits around $9 a month on the yearly plan in May 2026 and includes a native in-player Netflix overlay, the SRS, and the Migaku Academy course series. Migaku's browser extension can also create sentence flashcards automatically and export vocabulary directly to Anki, which is the kind of anki integration the open-source crowd has been building by hand for years. The Academy is a structured beginner-to-intermediate Japanese curriculum and is a real reason to pay for Migaku if you want a course alongside the mining workflow. Verify Migaku pricing on migaku.com before relying on it for a decision.
Language Reactor is the other name that comes up. Language Reactor is a free multi-language Netflix and YouTube overlay that provides dual-language subtitles and word definitions in real-time when hovering over Japanese text. It is not a JP-specialist tool and it does not include a spaced repetition system, so the loop ends with the dictionary lookup rather than the saved card. Useful for casual watching and for hearing native pronunciation in context, not designed for the full mining loop.
A note on the broader extraction landscape. Extracting Japanese vocabulary from YouTube can also involve downloading and parsing transcript text files, or using AI-driven language learning extensions that summarize lines into vocab lists. These are useful for power users who want batch workflows; the hover-save-review loop in this post stays inside the live video and is the simpler on-ramp.
Picking channels and shows for learning Japanese
Choosing the right japanese media matters more than choosing the right tool. The two filters worth applying are register and pace.
Register means how close the speaker's Japanese is to language a real person might use day to day. Slice-of-life TV show clips, casual interview channels, vlogs, talk-style variety shows, and many Japanese YouTuber long-form videos are conversational in a way that anime and period drama clips often are not. Pace means how fast the dialogue moves. Faster videos are not better practice; they are just harder, and difficulty without comprehension is not learning.
A practical pacing tip: YouTube lets you slow playback to 0.75x without distorting speech, which can help if the content feels too fast to capture new vocabulary. Use the speed control until your ear catches up, then return to 1.0x as comprehension grows.
For intermediate language learners, the practical advice is to start with content that feels slightly easier than your reading level. If you are around N3, a YouTube channel where you can understand Japanese in roughly seventy percent of each video is the sweet spot. You will find four to eight useful words per video, the watching stays fun, you will enjoy the channel enough to keep subscribing, and you will build the basics of an immersion habit you can actually maintain. The known words that show up repeatedly across episodes reinforce themselves at no extra review cost.
To track your progress over time, rewatch videos after some time has passed. Most language learners notice within two or three weeks that lines they once paused on now run by without confusion. That is the cleanest progress signal you will get, and it is what builds the confidence to tackle more difficult content later.
Thematic learning is also worth using on purpose. Thematic learning, where vocabulary is grouped by themes, is recommended for better memory retention when learning Japanese; the brain ties new words to a domain it already cares about. Pick one theme you are interested in (cooking, gaming, travel, business interviews, fashion, software) and binge channels in that theme for a few weeks. The repeated domain vocab compounds faster than the same hour spread across unrelated topics.
How many anki cards a day, and how to start sentence mining
There is no universal number, but the shape of the answer is the same for everyone. New cards generate roughly five future reviews each before they stabilize, so ten new cards today is fifty review slots over the next two weeks. Community recommendations on Anki and similar SRS tools typically suggest a ceiling of ten to thirty new cards a day depending on individual capacity. For a learner watching one Japanese YouTube video per night, four to eight new sentences per session is a sustainable starting point.
As language learners progress, they should reduce the number of new sentences they study each day rather than increase it. Immersion learning involves engaging with native materials such as TV shows, books, and online articles to enhance language acquisition, and you want review time to leave room for that. The process of immersion learning allows language learners to encounter new vocabulary in context, which aids in understanding and retention; if your review queue is two hours, you have no time for the immersion that makes the cards stick in the first place.
Daily, short study sessions are also more effective for building habits in learning Japanese vocabulary than infrequent long sessions. Ten minutes a day every day beats two hours every Saturday for the same total time, because the brain consolidates between sessions and because a daily habit is easier to keep than a weekly one.
How to start sentence mining if you have never done it: install immit, open one Japanese YouTube video with real Japanese captions, pause on the first sentence with a single unknown word, hover, save. Do it five times. You now have five cards. Later, review them. That is the whole on-ramp.
Where this loop goes next: Japanese web reading, same inbox
The YouTube loop above is one surface of a larger habit. The same hover-save-review pattern works on Japanese Twitter, on NHK News Web Easy, on Japanese Wikipedia, and on Aozora Bunko and Syosetu light novels. Manga is more format-dependent: if a manga page is rendered as raw images (the typical case for scanned releases), immit's popup dictionary cannot read the words because there is no selectable text on the page; if the same manga is converted to a text-layered format or read in a web reader that exposes selectable Japanese text, the hover-save-review loop works there too. Worth checking the format before you assume manga is covered.
Many language learning apps provide features that allow users to look up words instantly while reading native materials, facilitating a more immersive learning experience, and the immit popup dictionary is built for exactly that pattern across every Japanese web surface where text is selectable. The same apps often include functionalities for creating and managing flashcards, which can be tailored to individual learning needs and preferences.
immit is the same tool across every selectable-text surface, which means a sentence you mined from a YouTube interview tonight and a sentence you mined from an NHK News Web Easy article tomorrow live in the same review queue. The sustainability gain from one inbox across all your sources is the part that does not show up in feature checklists but shows up in whether you are still mining six months from now.
The pattern, applied across surfaces, is what most learners describe as the start of real fluency. You stop separating "study" from "watching things in Japanese" from "reading things in Japanese." It all becomes one habit with one inbox.
FAQ
Does immit work on Netflix Japanese subtitles?
Not currently. Netflix renders its subtitles inside a non-selectable player element, which means immit's popup dictionary cannot read them. If Netflix is your only immersion source, Migaku is the serious paid alternative (around $9 a month yearly in May 2026, with a native in-player overlay). immit's YouTube CC workflow is where the integrated lookup-and-SRS loop runs end to end today.
How do I mine Japanese vocabulary from YouTube videos with Japanese captions?
Install immit free in Chrome, open a Japanese YouTube video with real (not auto-generated) Japanese captions, pause on a line with one unknown word, hover, and click save. The full Japanese sentence is stored as a sentence card in immit's spaced repetition queue and resurfaces tomorrow for review. No AnkiConnect, no card template configuration, no Anki integration to wire up.
What is the best free way to mine vocabulary from Netflix in 2026?
The free path on Netflix is the asbplayer chain. asbplayer extracts Netflix subtitles into a separate selectable surface; Yomitan then handles the popup dictionary lookup on that surface (not on Netflix directly), and AnkiConnect saves the card to Anki. The setup is roughly half a day if you have not done it before. immit is not part of this chain because Netflix subtitle text is not selectable; immit covers YouTube CC and Japanese web reading instead.
How is the YouTube CC workflow different from Migaku's Netflix integration?
Migaku includes a native Netflix subtitle overlay, the SRS, and the Migaku Academy course in one paid subscription (around $9 a month yearly in May 2026). immit covers the YouTube CC mining loop and Japanese web reading free, with no account. If your immersion plan is Netflix-heavy, Migaku is built for that. If your plan is YouTube-heavy plus Japanese web reading, immit is the path. Verify Migaku pricing on migaku.com before deciding.
Can I use this workflow on auto-generated Japanese captions?
You can, but the results are mixed. Auto-generated captions miss kanji, drop punctuation, and stumble on regional accents, which makes the sentences less useful to mine. Stick to channels that include real Japanese captions whenever possible. The CC menu on each video shows whether the Japanese track is "auto-generated" or a real caption file.
What happens to my saved words when I switch from YouTube to reading on the web?
They stay in the same review queue. Whether you mined the word from a YouTube caption, an NHK News Web Easy article, a Japanese Twitter post, or a light novel on Aozora Bunko, the cards live in one immit inbox. All of this is free, and syncing across devices is free too with an optional account.
Is immit free for the YouTube mining workflow?
Yes. The popup dictionary, the one-click save to create a sentence card, the spaced repetition review queue, and offline use are all free with no account. If you run immit on a second machine, a free account syncs your cards across devices.
If you want to try the YouTube loop tonight, install immit free from the Chrome Web Store, open one Japanese video with real captions, pause on the first sentence with a word you do not know, and hover. The first card takes about thirty seconds. The next thousand take less.