How to Build an Anki Mining Deck for Japanese (No Yomitan Required)

The first time I tried to set up a mining deck, I spent a Saturday afternoon not learning a single Japanese word. I installed Anki, then the Japanese Support add-on, then AnkiConnect, then Yomitan, then a community note type someone swore by, then spent another hour figuring out why the cards came through with the fields in the wrong order. By the time the pipeline worked, I was too tired to read anything. The setup had eaten the study session.

If you have read a thread about sentence mining and walked away thinking "this sounds great, but the setup looks like a part-time job," you are not wrong, and you are not alone. The good news is that the actual idea behind a mining deck is simple, and the tooling around it has more than one shape. This guide covers what a mining deck is, how to use Anki for Japanese the standard way (Yomitan, AnkiConnect, card template and all), and how to run the same loop without assembling any of that. I am an American intermediate learner of Japanese, and I build immit with my wife who is a native speaker and our product designer. We made immit because the setup above is exactly the part neither of us wanted to do twice.

If you want to skip the build entirely and start sentence mining in the next minute, add immit free (no account). If you want to understand the whole picture first, keep reading.

Learning Japanese: where a mining deck fits

Learning Japanese is really several jobs stacked together. You learn the three writing systems first: hiragana, katakana, and kanji. You learn basic grammar, the particles and verb forms that hold a sentence together. And you learn vocabulary, thousands of words, over a long stretch of time. A mining deck is a Japanese vocabulary tool. It does not teach grammar and it does not replace a grammar resource, so if you are still working through basic grammar, keep a grammar guide running alongside.

It also is not the first thing an absolute beginner needs. If you do not yet know the writing systems, start there. The Tofugu hiragana and katakana guides teach both scripts in about two weeks, and that is a better use of week one than building a card pipeline. Mining makes sense once you can read native material slowly and you keep hitting unknown words faster than you can keep them. That is the moment this guide is written for.

Where does a mining deck sit relative to grammar study? Grammar gives you the structure; vocabulary fills it in. A grammar point you learned last month only becomes real when you meet it in fifty different sentences, and reading native material is how you meet those sentences. Grammar patterns settle in through exposure, not through a single explanation, so the value of language learning at this stage is in volume of input. Mining is what keeps the new words from that input instead of letting them slip away.

What a mining deck actually is

A mining deck is an Anki deck you fill with words you personally ran into while reading or watching Japanese, instead of words someone else picked for you. You meet an unknown word in native material, you make a card for it, and Anki schedules that card for review using its spaced repetition algorithm. That is the entire idea.

Sentence mining is the practice that feeds the deck. You extract a sentence that contains an unknown word, save it, and review it on a spaced schedule until the word is yours. The cards built this way are sometimes called targeted sentence cards: each one targets one new word inside a sentence you can already mostly read. The reason it works is context. When you learn a word inside a sentence you actually read, your brain has a hook to hang the meaning on. A single word floating alone is harder to keep than the same word sitting in a sentence about something you cared about.

The distinction that matters is this. A prebuilt deck teaches you the most common words in a fixed order. A mining deck teaches you the certain words that are actually blocking your reading right now. As you read more, new sentences bring new vocabulary and new grammar you have not seen, and the deck grows out of exactly that. Both approaches have a place, and most learners use both at different stages.

Why mine your own deck instead of grabbing the best Anki deck

There are excellent prebuilt decks, and ignoring them would be silly. Kaishi 1.5k covers roughly 1,500 common Japanese words with audio and is the standard modern starter deck. Tango N5 covers about 1,000 words aimed at JLPT N5 preparation, which is useful since the N5 level tests around 800 vocabulary words. Recognition RTK teaches around 2,200 kanji meanings using English keywords. Core 6000 extends vocabulary coverage to about 6,000 common words. If you are early and you want fast coverage of common words, the best deck for you is whichever one matches your level. (Verify deck contents and word counts on AnkiWeb before you commit, since community decks get revised.)

The problem prebuilt decks run into is the plateau. Conversational fluency in Japanese is usually estimated at somewhere between 5,000 and 10,000 words, and the common-word decks stop well before the top of that range. Past the first few thousand words, the useful words you need stop being "common words everyone learns" and start being "the words in the specific manga, novels, and videos you read." Nobody has built a deck for your reading list. That is the point where mining your own deck stops being optional and starts being the only way forward.

There is also a retention difference. Learning words from your own immersion content tends to stick better than mass-importing a deck, because every card is tied to a moment you actually experienced. The word is not abstract. You remember reading manga on a Tuesday and hitting a wall on that exact word. For many learners, that context is the difference between internalizing new vocabulary in weeks instead of months.

How other tools handle the same job

Anki is not the only place people make flashcards, and it is worth knowing where it sits among other language learning tools. Migaku lets you create flashcards directly from Japanese media like Netflix and YouTube, which is its main draw and a real reason some learners pay for it. WaniKani offers a structured kanji learning system that runs on its own SRS, so it covers the kanji job rather than the open-ended vocabulary job a mining deck does. Quizlet provides mobile-friendly flashcards that are easy to start with but light on the spaced repetition Anki is built around. LingoDeer focuses on sentence drills with grammar explanations, so it leans toward the grammar side. KanjiDamage uses memorable mnemonics for kanji learning, which can help when a character refuses to stick.

None of these is the wrong tool. They are designed for different jobs, and serious language learners often run two or three at once. WaniKani owns structured kanji curriculum, and if a guided course through the kanji is what you want, it does that well. A mining deck, in Anki or in immit, is for the part those tools do not cover: the long tail of vocabulary in your target language that only shows up in the specific things you choose to read.

The standard build: how to use Anki for Japanese with Yomitan and AnkiConnect

Here is the honest version of the manual setup, so you know what you are choosing between. This is a real, powerful, free workflow, and plenty of learners love it. It is also the part that ate my Saturday.

First, download Anki desktop from apps.ankiweb.net. It is free. Install the Japanese Support add-on, which adds reading generation and a few Japanese-specific features. Install AnkiConnect, the add-on that lets a browser extension talk to Anki and create cards automatically. Then install Yomitan, the popup dictionary, and configure its Anki integration: you tell Yomitan which deck to send cards to, which note type to use, and which dictionary field maps to which card field.

That note type step is where most people stall. You have to pick or build a card template, then map the Yomitan fields (word, reading, definition, sentence, audio) onto the front and back of that template. Get the mapping wrong and your Anki cards arrive blank or scrambled, and you are back in the settings menu instead of reading. None of this is hard once it works. It is just a lot of moving parts to align before you make card number one.

When it is set up, the loop is genuinely good: hover a word in Yomitan, press a hotkey, and a formatted card lands in Anki. The cost is the thirty-plus minutes up front and the occasional afternoon of debugging when a Chrome update or an add-on version shifts something.

The card template that works

Whatever path you take, a good Japanese word card has a predictable shape. The front carries the prompt. The back carries the answer: the reading in kana, the meaning, the part of speech, and ideally the example sentence the word came from plus audio.

A few things make cards stick. Example sentences give the word context, which is the whole reason mining beats isolated word lists. Audio clips improve pronunciation and listening, so you are not silently guessing how a word sounds. Visual mnemonics can help for stubborn words, especially kanji-heavy ones, though they are optional and easy to overdo. The one habit worth keeping no matter what: short sentences. When you find sentences to mine, pick ones short enough to read in one breath. Useful sentences for a card are the ones that show the word doing its job without a wall of extra phrases around it. A mining card built on a fifteen-word sentence is a card you will come to dread.

What you are avoiding is the overstuffed card. Front with the word, back with reading, meaning, one short example, and audio. That is enough. Everything past that is weight you carry through every future review.

How many new cards per day, and how to handle reviews

This is where most mining decks quietly die, so it matters more than the template. The failure mode is adding too many new cards, watching the daily review pile balloon, and quitting.

In Anki, set a sane new-cards-per-day limit in the deck settings. For mining, somewhere around 10 to 20 new cards a day is a sustainable ceiling for most people, and 10 is a perfectly good number if you also have a job. Review daily, not in occasional marathons. Fifteen to twenty minutes a day beats two hours on Sunday, because spaced repetition only works if the spacing is respected. This is how Anki handles scheduling: Anki's algorithm sets each card's next review for just before you would forget it, and that timing breaks if you skip days and then binge on review cards.

When you grade a card, prefer "Good" over reflexively hitting "Easy." Easy pushes the next review far into the future, and on a word you barely know that is how you end up failing it weeks later with no idea why. Let the algorithm do its job at the normal pace. The point of Anki study is consistency, not speed.

Manage your leeches. A leech is a card you keep failing no matter how many times it comes around. Anki can suspend these automatically so they stop stealing time from cards that are actually moving. Suspending a stubborn card is not giving up. It is parking the word until you meet it again in real reading, which is usually when it finally clicks.

Done this way, a mining deck can carry you a long way. Many learners use Anki to retain well over 10,000 words across years of daily reviews. The trick is never the algorithm. It is keeping the daily load small enough that you actually show up.

The no-Yomitan path: skip the chain entirely

Here is the part the title promised. You can run the exact same mining loop, meet a word, save it, review it on a spaced schedule, without installing Yomitan, AnkiConnect, a Japanese Support add-on, or a custom note type. That is what we built immit to do.

immit is a popup Japanese dictionary with a built-in 8-stage spaced repetition system. You install it as a Chrome extension and a desktop app. You hover any Japanese word on any web page and the reading, part of speech, definition, an example sentence, and pronunciation audio appear in about a tenth of a second. One click on the bookmark icon saves the word to your personal deck. You can also use the built-in search to look up a word you are thinking of rather than one you are hovering over. There is no AnkiConnect to wire up and no field mapping to get wrong, because the lookup, the save, and the review all live in the same tool. Lookup, save, and review in one place is the whole substitution.

The card immit makes for you on save uses a sensible default: the word on the front, and the part of speech, definition, example sentence, and pronunciation audio on the back. You review in two modes. Flip mode shows you the word and you reveal the answer, then mark it easy or difficult. Type mode asks you to type the answer and checks it. There is no deck-settings panel to configure and no new-cards-per-day dial, because new cards enter your queue as you save them while reading. Your pace is set by how much you read, not by a number you picked in a menu. If a week gets heavy, save fewer words and let the queue settle.

What you get in exchange is no setup. The first lookup happens within thirty seconds of install, the dictionary works offline because the data lives on your device, and there is no account required to look up, save, and review. If that trade sounds right, add immit free and the desktop app, and start mining on the next page you read.

What you keep and what changes when you drop the manual stack

Switching the tool does not change the method. You are still sentence mining: reading native material, saving the words that block you, reviewing on a spaced schedule. The spaced repetition principle that retains those 5,000 to 10,000 words is doing the same work either way.

What changes is everything around the method. You lose the deep customization of the Anki plus Yomitan stack: custom note types, dictionary imports, fine-grained scheduling tweaks, the full sentence-card setup. If you enjoy that configuration as part of the hobby, you should keep it, and this post is not trying to talk you out of a workflow you love. What you gain is the thirty minutes of setup back, plus every future afternoon you would have spent debugging an add-on after a browser update. For a lot of learners, that trade is the difference between a mining habit that survives and one that dies in the settings menu.

FAQ

What is a mining deck in Anki?

A mining deck is an Anki deck you fill with words you personally encountered while reading or watching Japanese, rather than a fixed list someone else assembled. You make a card the moment you hit an unknown word in native material, and Anki schedules it for spaced review. The deck grows from your own reading, so it teaches the words actually blocking you right now.

Is Anki good for learning Japanese?

Anki is good for the vocabulary and kanji part of learning Japanese, and it is the community standard for sentence mining because its spaced repetition is strong and it is free. It is less good at being easy to set up, and it does not teach grammar. Most learners pair Anki with a grammar resource and either a popup dictionary like Yomitan or an all-in-one tool like immit.

Do I need Yomitan to build an Anki mining deck for Japanese?

No. Yomitan is the popular popup dictionary that feeds cards into Anki through AnkiConnect, and it is a strong free tool, but it is not the only way. You can build cards manually, use a different popup dictionary, or use an all-in-one tool like immit that combines lookup, save, and spaced repetition so there is no separate dictionary and no AnkiConnect bridge to set up.

What is AnkiConnect, and can I skip it?

AnkiConnect is an Anki add-on that lets a browser extension like Yomitan create cards in Anki automatically. It is the glue in the standard mining setup. You can skip it if you are not using Anki as your review engine. immit, for example, has its own built-in spaced repetition, so there is nothing to connect.

Should I mine my own sentences or use a prebuilt deck like Kaishi 1.5k?

Both, at different stages. Early on, a prebuilt deck like Kaishi 1.5k or Tango N5 gives you fast coverage of the most common words. Once you are past the first few thousand words, the vocabulary you need is specific to what you read, and no prebuilt deck covers that. Mining your own deck takes over from there.

What fields should a Japanese mining card have?

Keep it lean. The front holds the word. The back holds the reading in kana, the meaning, the part of speech, a short example sentence the word came from, and audio if you can get it. Resist adding more. An overstuffed card is one you will dread reviewing.

How many new cards a day should I mine without burning out?

Around 10 to 20 new cards a day is sustainable for most learners, and 10 is fine if you are busy. Review daily for 15 to 20 minutes rather than in long weekend sessions, since spaced repetition depends on consistent timing. The most common reason mining decks fail is too many new cards at once, not too few.

Can I run a mining loop without Anki at all?

Yes. Anki is one review engine, not the only one. immit runs the full loop, hover to look up, one click to save, built-in spaced repetition to review, without Anki, AnkiConnect, or any add-ons. The method is the same; the tool is simpler.

Is immit free, and does it work offline?

immit has a free tier that covers lookup, saving words, and spaced repetition review with no account required, and the dictionary works offline because the data is stored on your device. Pro is $9 a month or $108 a year and adds multi-device sync, cloud backup of your flashcards, and dark mode.

Start with the next thing you read

The hardest part of sentence mining was never the reviewing. It was the setup standing between you and the first card. If the standard Anki and Yomitan build is the workflow you want, this guide gave you the honest version of it, template, pacing, and all. If the setup is the reason you keep bouncing off mining, that is the exact problem immit removes: add immit free and the desktop app, open the next thing you actually want to read, and save the first word that stops you. The deck builds itself from there.