Affiliate disclosure: some of the links below are affiliate links, which means I may earn a commission if you buy through them, at no extra cost to you. I want to be blunt about what that does and does not buy: the single app I recommend most often in this article pays me nothing at all, because it has no affiliate program. I have flagged which products can pay me and which cannot, in every section. If a recommendation ever reads as suspiciously convenient, check that flag.
I want to start by telling you something about the articles you have probably already read.
Search for the best app to learn Japanese and look closely at who published each result. One is published by italki. One by Clozemaster. One by Migaku. One by Lingrow. One by an app called noun.town. Each of those companies sells a Japanese learning product, and each of them has written a roundup of the best Japanese learning products. You can guess how those rankings turn out.
That is not a conspiracy, it is just marketing, and it works because from the outside every app looks roughly the same: a friendly interface, a streak counter, some cheerful audio. What none of those articles do — and I mean none of the ones I read while researching this — is answer the question that actually matters to an adult learner.
Is the Japanese in this thing any good?
I am a native speaker. I live in Tokyo. I am not qualified to tell you which app has the best gamification loop, and I am frankly not that interested. What I can tell you is when an app is teaching you sentences that will make a Japanese person’s face do something polite and strange. That is the gap I am writing into.
The problem nobody names
Here is the thing that happens to adults who learn Japanese from apps, and it happens with remarkable consistency.
You spend eight months building a streak. You can read hiragana comfortably, you know a few hundred words, you have absorbed a surprising amount of kanji. Then you get into an actual conversation with an actual Japanese person and produce almost nothing. Not because you lack vocabulary. Because the app never once asked you to make a sentence — it asked you to recognise one, tap it, and receive a chime.
There is a second failure that is more specific and, from my side of the conversation, more noticeable. The app taught you Japanese that is technically correct and socially wrong. Textbook Japanese defaults to a register — the polite です／ます form — and then stops, so learners arrive fluent in a voice that sounds like a train announcement. Or worse, an app drilled you on casual forms without ever explaining who you are allowed to use them with, and now you are speaking to your girlfriend’s father like he is your college roommate.
In my experience, that second problem is the one that actually stalls people, and it is invisible from inside the app. Your streak does not know. Duolingo’s owl does not know. You find out in a room, in real time, from a stranger’s face.
So this article is organised around one question: which of these products will leave you able to say something a Japanese person would actually say?
Who this article is not for
I would rather lose you here than waste your afternoon.
If you want to learn Japanese casually, for fun, with no particular endpoint — stop reading and use Duolingo. I am serious. It is free, it is genuinely pleasant, and if your honest goal is fifteen enjoyable minutes a day, the fact that it will not make you conversational is not a flaw. It is a mismatch between the tool and what you actually want. Everything below assumes you want to reach real usage, and that assumption makes the recommendations more demanding and more expensive than you need.
If you are learning Japanese for an upcoming trip in under three months — none of this is the right investment. Learn to read hiragana and katakana, learn about thirty set phrases, and spend the rest of your energy on logistics. Japan is extraordinarily navigable for visitors who speak no Japanese. Three months of app study will not change your trip, and I would rather you spent the money on a better dinner.
If you are a heritage speaker or you already understand spoken Japanese from family — your bottleneck is literacy and formal register, not beginner material, and almost every product below will waste your time on things you already know. You want kanji tools and reading practice, so skip ahead to the WaniKani and Anki sections and ignore the rest. The other thing worth your attention is honorific language, which is where fluent-sounding speakers get caught most often.
If you are hoping one app will be enough — it will not be, and any article that tells you otherwise is selling you the one app. Japanese splits cleanly into skills that different tools handle badly and well. Budget for two, used together.
How I judged these
Three criteria. I applied them in this order, and the order matters.
1. Does it make you produce Japanese, or only recognise it?
Recognition is easy and it feels like progress. You see 食べる, you remember it means “to eat,” dopamine arrives. Production is the hard part: constructing a sentence from nothing, choosing the verb form, choosing the politeness level, and getting it out of your mouth before the moment passes.
Japanese counters are the same problem in miniature.
The gap between those two abilities is enormous, and most apps are recognition machines wearing a production costume. Multiple choice is recognition. Tapping word tiles into order is recognition with extra steps. Typing a full sentence from an English prompt is production. Speaking out loud into a pause is production. I weighted production heavily, because it is the thing that fails you in the room.
2. Is the Japanese natural, and does it teach register?
This is the criterion I am uniquely positioned to apply, so let me be precise about what I mean.
Natural does not mean casual. It means a Japanese person in that situation would plausibly say that sentence. A lot of learning material fails this in a specific way: it is grammatically flawless and contextually bizarre, built to demonstrate a grammar point rather than to communicate. Sentences like “This is a pen, that is not a pen” survive in teaching material long after they stopped resembling human speech. The wider version of this problem — what actually separates textbook Japanese from the language people speak — turned out to be large enough to need its own article, and the biggest piece of it is not vocabulary at all.
Writing is where you can practise this deliberately, which I cover in my guide to Japanese writing practice apps.
Register is the other half. Japanese encodes your relationship to the listener in the verb itself, which means there is no neutral option — every sentence you speak takes a position on who you are to the person in front of you. A product that teaches you forms without teaching you when to deploy them has handed you a loaded thing without the safety instructions. I marked products down hard for this, and it separated them more than any other criterion. I have since given this criterion an article of its own, on which products actually teach casual Japanese and which leave you stuck in polite form.
3. Does it survive month six?
Almost anything works for three weeks. The question is whether the thing still functions when the novelty is gone, when you have hit the intermediate plateau, and when your daily session has become a chore you do on the train.
Practically this means: is there enough content to keep going, does the difficulty actually keep climbing, and does the review system handle the fact that you will inevitably miss a week? The apps that fail here are the ones that were designed to be delightful rather than to be finished.
The comparison
Can it pay me? = whether the product runs an affiliate programme that could pay me a commission if you buy through my link. It is not a claim that the product is paying me. Only JapanesePod101 has approved me, and its entry is the one tracked link on this page. I would rather you saw this next to the ratings than buried at the bottom.
| Product | What it is actually for | Makes you produce? | Natural Japanese / register | Survives month 6? | Can it pay me? |
|---|---|---|---|---|---|
| WaniKani | Kanji and vocabulary, nothing else | Typing readings and meanings — production of a narrow kind | Excellent vocabulary, no register teaching (not its job) | ★★★★★ Built for the long haul | No — no affiliate program exists |
| JapanesePod101 | Listening and dialogue at every level | Weak. Mostly listening | Strong. Real conversational dialogue, register explained | ★★★★☆ Enormous library | Yes |
| Rocket Japanese | Structured course with voice practice | Yes. Voice recording and full-sentence recall | Good. Deliberately teaches politeness levels | ★★★☆☆ Finite course | Yes — I am not in it |
| Pimsleur Japanese | Speaking and listening, audio only | Yes. Forces spoken output constantly | Good but conservative, polite register only | ★★★☆☆ Runs out | Yes |
| Bunpro | Grammar, via spaced repetition | Yes. Type the grammar into real sentences | Very good. Written by people who care | ★★★★☆ Grammar-only by design | No — points, not cash |
| Anki | Whatever you make it | Depends entirely on your deck | Depends on your deck | ★★★★★ Runs forever | No — free and open source |
| Rosetta Stone | Immersion-style beginner course | Some, but shallow | Weakest here for Japanese specifically | ★★☆☆☆ | Yes, and I still do not recommend it |
What I actually checked, and what I did not
A note on method, because that table is only worth as much as the testing behind it.
I went through JapanesePod101’s beginner dialogues on the free tier with one question in mind: would I flinch? That is the whole test I have been describing — not whether a sentence is grammatical, but whether a person in that situation would say it.
On the sentences themselves, nothing flinched. The dialogue Japanese was not stiff. I found no museum pieces, no lines built to demonstrate a grammar point rather than to communicate. And the thing I was most ready to catch them on, the movement between です／ます and plain form, was handled naturally. Characters shift register the way people actually shift register, instead of parking in polite forms for the entire course the way beginner material usually does.
Then I found the real problem, and it is not the one I went looking for. Some of this material is built around situations you will almost never be in. The Japanese inside them is perfectly good Japanese — a native speaker would say those words in that room — but you will not be in that room. That is a different failure from the one I have been criticising all article, and it is worth naming precisely, because the fix is different too. Duolingo gives you unusable sentences. This gives you usable sentences attached to unusable occasions.
Here is one I was actually served:
頭を守るためにヘルメットをかぶりなさい。
Atama o mamoru tame ni herumetto o kaburinasai.
Put a helmet on to protect your head.
Nothing about that is wrong. It is correct, it is idiomatic, and a Japanese person could say it. Now ask yourself when you are going to say it.
And there is a second thing in that sentence which a beginner cannot see. The verb ends in なさい, and なさい only travels downward. A parent uses it on a child. A teacher uses it on a class. An exam paper uses it on a student. It is the register of instruction, and it carries an assumption about who is above whom that is baked into the ending itself.
So picture what happens if that form goes in through repetition, the way drilled sentences do, and comes back out eight months later on a colleague at a construction site or a friend on a bicycle. You have not made a grammar mistake. Nothing you said was incorrect. You have told a grown adult to put their helmet on in the voice a mother uses on a seven-year-old, and the face you get back will not tell you why. That is precisely the failure this whole article is about, and I found it sitting inside the product I ranked second.
That is a curation problem rather than a language problem, and with a library this size it is the failure you would predict: fifteen years of lessons, produced continuously, with far more pressure to keep publishing than to prune. Volume is this product’s great asset and the source of its only serious flaw.
So use it deliberately, not linearly. Do not start at lesson one and work forward, which is how the interface invites you to use it and how you will end up rehearsing conversations you will never have. Pick by situation instead. Search the library for the thing you actually need to be able to do this month, study that dialogue properly, and skip freely. The register commentary is the asset. The running order is not.
Now the part I owe you, because this is a product that pays me. I checked beginner dialogues on the free tier. I did not audit a fifteen-year back catalogue, so the verdict I will defend is narrow: the Japanese in the dialogues is real, the register moves, and the situation selection is padded. I earn a commission on this one, and it still has the flaw above, which I would rather you heard from me than discovered in month three.
What I did not test. Pimsleur and Rosetta Stone both put their trials behind a credit card that renews automatically into a paid subscription. I did not hand a card to either one, so my read on those two comes from their published method and materials rather than from use. I would rather say that than let you assume otherwise. It does mean my sharpest claim about Rosetta Stone — that immersion without explanation is a bad fit for a language which encodes social relationship in the verb itself — is an argument about method, which I will gladly make, and not a report from inside the product.
The products, one by one
WaniKani — the best thing here, and it cannot pay me
What it does: teaches you roughly 2,000 kanji and 6,000 vocabulary words through spaced repetition, in a fixed order that builds from radicals upward. That is the entire product. It does not teach grammar, it does not teach conversation, it will not help you order lunch.
Why it works: the ordering is the product. Kanji are not arbitrary shapes, they are compositional, and WaniKani sequences them so that every new character is built from pieces you already know. It also insists you type the reading, which is a small thing that matters enormously — you cannot bluff your way through with vague recognition.
The vocabulary choices are genuinely good. When I look at their word lists I see words people actually use, not the museum pieces that fill a lot of textbooks.
Pros: best-in-class kanji sequencing; forces recall rather than recognition; the community is unusually helpful; lifetime pricing exists and pays for itself if you stick with it.
Cons: the pacing is enforced and you cannot rush it, which is correct and also maddening; the mnemonics are aggressively silly and some people bounce off them; it teaches you zero grammar and zero conversation, so it is half of a study plan, never the whole one.
Who should buy it: anyone serious about reading Japanese. If you plan to study for more than a year, this is the highest-value thing on this page.
A note on my incentives: WaniKani has no affiliate program. Learners have been asking Tofugu for one for years and it does not exist. I get nothing from this recommendation, which is exactly why I feel comfortable making it the strongest one in the article.
JapanesePod101 — enormous, uneven, and the best listening resource here
Updated 22 August 2026: I was approved for the JapanesePod101 affiliate programme on that date, so the link in this heading is a tracked one and it pays me 25 percent. It is the only tracked link on this page. Nothing in the ranking has moved, and the two products above it in my list still pay me nothing.
What it does: a vast library of audio and video lessons, structured as dialogues with breakdowns, from absolute beginner through advanced. Subscription based, with tiers that unlock more study tools.
Why it works: the dialogues are the real asset. They are performed by native speakers at natural speed with a slowed version alongside, and — this is the part I care about — the hosts routinely stop to explain why a character used that form to that person. That register commentary is rare, and it is the exact thing most apps skip.
Pros: the library is genuinely enormous and you will not exhaust it; audio quality is high; register and cultural context are explained rather than assumed; the cookie on their affiliate program never expires, which I mention only so you know I have a long-term interest in you clicking it.
Cons: the marketing is relentless and the upsells are constant, which is the most common complaint and a fair one; quality varies across a fifteen-year back catalogue and some older lessons feel dated; it is a listening resource wearing the costume of a complete course — you will absorb a great deal and produce very little unless you pair it with something else.
Who should buy it: learners whose listening is the weak link, and anyone who wants to understand why Japanese people say things, not just what they say.
Before you pay, one thing worth knowing: every JapanesePod101 subscription carries a 60-day money-back guarantee. It is on their own pricing page, in the FAQ at the foot of it — your purchase comes with a 60-day money-back guarantee – no hassles, no questions asked. I checked it again on 27 August 2026. Two caveats I owe you: you claim it by emailing them rather than by clicking a button, and a refund is deducted back off my commission, so the 25 percent above is money I lose if this turns out to be wrong for you. Take the refund anyway.
Rocket Japanese — the most complete single course
What it does: a structured, level-based course combining audio lessons, written grammar explanations, and voice recording that compares your pronunciation against a native speaker’s.
Why it works: it is the closest thing here to an actual curriculum with a beginning and an end. The voice recording feature sounds like a gimmick and mostly is, but the act of being forced to say the sentence out loud before moving on does real work.
Rocket also handles politeness levels more deliberately than most, introducing register as a concept early rather than dumping casual forms on you at level three with no explanation.
Pros: genuinely structured; forces spoken output; one-time purchase rather than an indefinite subscription, which some people strongly prefer; teaches register explicitly.
Cons: the interface looks like it was designed a while ago; the price is high up front; it is finite — you will finish it, and then you need something else; the voice comparison technology is not as clever as the marketing implies.
Who should buy it: adults who want one clear path to follow and dislike assembling their own curriculum, especially those who hate subscriptions.
Pimsleur Japanese — the only one that really makes you talk
What it does: audio-only lessons, thirty minutes each, built around a strict spaced-recall method. A voice asks you to say something in Japanese, leaves a silence, and you fill it. No screen, no tapping, no streak.
Why it works: the silence. It is the only product here that puts you under mild pressure to produce spoken Japanese from nothing, repeatedly, in every single session. That is uncomfortable and it is precisely the skill that fails in real conversation. In my experience, the learners who can actually hold a conversation are the ones who practised talking, and most apps quietly let you avoid it.
Pros: unmatched for spoken production; works entirely hands-free, so it fits commutes and walks; users report consistently that pronunciation and rhythm come out noticeably better.
Cons: expensive; no reading or writing at all, so it does nothing for kanji; the vocabulary is business-traveller conservative and stays in polite register throughout, meaning you will sound courteous and slightly formal everywhere; the method is repetitive by design and some people find it unbearable.
Who should buy it: anyone whose goal is speaking, especially if you have already studied Japanese and can read but freeze when you have to talk. I have written separately about why understanding Japanese and speaking it are two different problems. The method suits some people enormously and others not at all, so try it before committing — but note that the trial asks for a card up front and rolls into a paid subscription by itself. Put the cancellation date in your calendar on the day you sign up.
Bunpro — grammar, done properly, for very little money
What it does: spaced repetition for grammar points rather than vocabulary. It gives you an English sentence and asks you to supply the Japanese grammar structure, then reviews it on a schedule.
Why it works: grammar is exactly the thing that fits spaced repetition and almost nobody applies it there. It also links out to explanations across multiple sources for each point, which respects your intelligence.
Pros: cheap — $5 per month or $120 for lifetime access as of August 2026; the grammar explanations are careful and the example sentences are natural; pairs perfectly with WaniKani, since between them they cover kanji, vocabulary, and grammar.
How it compares with the other grammar tools is a separate guide. Cons: grammar only, deliberately; the interface is functional rather than lovely; it assumes you can already read kana and will be rough going if you cannot.
Who should buy it: anyone using WaniKani. The two together are close to a complete written-Japanese curriculum for under the price of one premium subscription.
A note on my incentives: Bunpro’s referral programme pays in points and shop items, not cash. I earn nothing from this.
Anki — free, ugly, and the last flashcard app you will ever need
What it does: open source spaced repetition. Empty by default. You either build decks or download ones other people made.
Why it works: it is infinitely flexible and it will still exist in ten years. Every serious learner I know ends up here eventually, usually after outgrowing something friendlier.
Pros: free on desktop and Android; total control; the mining workflow — pulling sentences from shows and books you are actually reading into your own deck — is the single best intermediate study method I know of. If the shows you mine from are anime, read this first, because a share of what you collect will be character speech rather than Japanese anybody says.
Cons: the interface is genuinely unpleasant; the iOS app costs real money; it requires you to design your own study plan, which is a skill beginners do not have yet and should not have to acquire; downloaded decks are wildly variable in quality and some contain Japanese I would not want you repeating.
Who should buy it: intermediate learners, and beginners only if you enjoy building systems. Free, so try it early and abandon it without guilt if it is not for you.
Rosetta Stone — the one I would skip
What it does: picture-based immersion learning with no translation, the same method it has applied to every language it sells.
Why it does not work well for Japanese: the method’s core premise is that you can infer meaning from images without explanation. That premise strains for any language and it breaks for Japanese, because the things that make Japanese hard are not nouns. You cannot infer politeness level from a photograph. You cannot infer particle usage from a picture of a woman holding an apple. The features Japanese encodes most heavily are precisely the ones that images cannot show.
Pros: polished; the speech recognition is generally well reviewed; the brand recognition means it is often bought as a gift.
Cons: expensive relative to what it delivers; weak on kanji; teaches essentially no register; you will finish a lot of exercises and still be unable to construct a sentence about anything that happened yesterday.
A note on my incentives: Rosetta Stone has a perfectly good affiliate program and I would earn a commission if you bought it. I am telling you not to.
What learners have asked me, and what I could not answer
I am not a Japanese teacher. I am simply a native speaker who lives here, which means people ask me things. Three questions have come up more than any others, and I want to be honest about how I did with them, because my failure to answer is the whole reason I rate these products the way I do.
“Why does one kanji have several different readings?”
I can read them. I cannot explain them. I acquired the on and kun split the way every Japanese person does, which is to say by twelve years of school and a lifetime of reading, without ever being handed a rule. When someone asks me why the character in question is read one way here and another way there, the honest answer is that I know which one is right and I do not know why.
This is exactly why I put WaniKani at the top of this article. A native speaker cannot transmit this. There is no shortcut I can offer you across the table in a cafe. What you need is a system that feeds you readings in a deliberate order and tests you until they stick, which is a thing software is genuinely good at and a thing I am useless for. The people who built WaniKani solved a problem that I, as a native speaker, cannot even describe properly.
“Why do you need three writing systems at once?”
I have been asked this more than once and I have never given a satisfying answer. The functional answer is that Japanese is written without spaces, so the shift between scripts is doing work that spaces do in English. That is true and it still does not make the burden feel reasonable to the person asking.
What matters for this article is the consequence. Any product that lets you feel you are progressing without ever confronting the writing system is deferring the bill, not cancelling it. That is my central objection to audio-only and picture-only approaches. Pimsleur will make you speak, and I recommend it for that, but it will not get you into the writing system. Rosetta Stone’s picture method cannot even show you the distinction. If you intend to read anything at all, you need something in your stack that treats the script as the main event rather than an accessory.
“Chopsticks and bridge sound identical. How do you tell them apart?”
They are not identical. Chopsticks drops after the first syllable, bridge rises and drops after the second, and the word for edge stays flat. I produce all three correctly and without effort, and the first time I was asked to explain the difference I could only say them out loud and hope the difference was audible.
That is pitch accent, and it is the largest shared blind spot on this page. I want to be precise about it, though, because when I went back and checked properly the picture was not as clean as I first assumed.
Two of these products do acknowledge it. JapanesePod101 has a pronunciation series with a lesson given over to pitch accent, which lays out the basic rules and drills a handful of words. Bunpro displays the pitch pattern on its vocabulary entries. On WaniKani it exists only if you install a userscript written by a community member, which is another way of saying it is not in the product. For the rest I found nothing.
But look at where those two exceptions sit. One is a single short lesson parked inside a pronunciation guide, off the main path, which you will meet once and never again. The other is a piece of data printed beside a word you are being tested on for something else entirely. In neither case is pitch part of what you are taught, drilled, or graded on. You can finish either product in full, answering every question correctly, and never once be required to produce the right pitch. That is the actual gap. Calling it a flat absence would have been the easier sentence to write, and the wrong one.
I am not going to pretend this changes my rankings, because a product cannot be marked down for a gap the entire category shares. But it does change my advice. If you get to an intermediate level using anything on this list and people still react to your Japanese as slightly off while your grammar is clean, this is very likely why, and you will have to go outside these products to fix it. I went and looked at what is outside them, and wrote up the tools that do teach pitch accent, none of which are on this page.
The pattern in all three: I can produce Japanese perfectly and describe it badly. That gap is the thing to hold onto when you read reviews written by learners, who often have the opposite profile. They can describe the rule and cannot yet produce it. Both of us are half blind, and a review written by either of us alone should be read with that in mind.
What to buy, based on what you actually want
“I want to read manga, novels, or the internet.” WaniKani plus Bunpro. This is the strongest pairing on the page and it is not close. Roughly a year in, you will be slowly reading real material. Neither pays me. When you get there, the reading tools that remove the lookup problem matter more than more study, because dictionary friction is what stops most people rather than the kanji themselves.
“I want to have conversations.” Pimsleur for spoken production, plus JapanesePod101 for listening and register. Add WaniKani only if literacy also matters to you — if it genuinely does not, skip kanji entirely and feel no guilt about it.
“I want one thing to follow and I do not want to assemble a plan.” Rocket Japanese. It is the most complete single product here. Accept that you will need something else afterwards.
“I am on a tight budget.” Anki plus Bunpro plus free grammar resources, total cost about $5 a month. This is slower and requires more self-direction, and it absolutely works.
“I am preparing for the JLPT.” WaniKani and Bunpro, both of which are organised in a way that maps onto the test. Nothing else here is built for it. I have since written a full guide to what the JLPT actually measures and which tools match it, and the short version is that the exam tests recognition and never production.
“I am moving to Japan for a job.” Start with the fixed formulas and the 内/外 rule rather than with grammar. My guide to business Japanese courses covers what is worth paying for.
“I tried Duolingo for a year and feel stuck.” This is the most common message I get, so let me answer it directly. You are not stuck because you lack discipline — you spent a year on recognition and never practised production. Go straight to Pimsleur or Rocket Japanese. The unstick is output, and it will feel bad for about two weeks.
The bottom line
If you take one thing from this article, take the criterion rather than the ranking: judge a Japanese course by whether it makes you produce Japanese, and by whether the Japanese it teaches is something a person would say. Most of them fail both tests while feeling excellent, and the feeling is the trap.
My honest ranking, in order, with commissions marked so you can weigh them:
- WaniKani for kanji — pays me nothing
- Bunpro for grammar — pays me nothing
- Pimsleur for speaking — could pay me, but I am not in that programme
- JapanesePod101 for listening and register — pays me 25 percent, the one on this list that does
- Rocket Japanese if you want one structured path — could pay me, but I am not in that programme
- Anki once you are intermediate — pays me nothing
- Rosetta Stone — could pay me and I am not in that programme, and I would still skip it
Two of the top two pay me nothing. That is not me being noble, it is just what happens when you rank on merit and then check the commission afterwards instead of before.
If you are starting today: get WaniKani and Bunpro and give them ninety days before you evaluate anything. If speaking is the actual goal, start Pimsleur’s free trial this week — the discomfort of the silence is the whole point, and you will know within three lessons whether the method is for you.
Have you hit the wall I described — a year of study and nothing comes out of your mouth? Tell me what you were using and I will tell you honestly whether the tool or the method was the problem.
— Momo Kurata, Tokyo