How to Build a Chinese Vocabulary That Actually Sticks (Free)

Most Mandarin learners have been here: you sit down to study vocabulary, you open a flashcard app, you grind through 30 cards, you close the app, and three days later you can barely remember ten of them. You start to wonder whether Mandarin is unlearnable, or whether your brain is broken. Neither is true. The problem is the vocabulary system, not the learner.

This guide explains how to build a Chinese vocabulary that actually sticks, using free tools, with no Anki setup and no subscription. It is the same system we use to design ReadDeck, the offline Mandarin reader we ship, and the same system the spaced repetition research community has converged on in 2026: read real sentences, mark the words you do not recognise, and let a modern scheduler (FSRS) handle the review.

If you have ever abandoned a flashcard app after a fortnight, read on. The fix is simpler than you think.

Why most vocabulary apps fail

The reason Mandarin vocabulary feels harder than, say, Spanish vocabulary is not the characters. It is the forgetting curve. Hermann Ebbinghaus worked out the curve in 1885 and it still holds: without review, you forget roughly 70 percent of new vocabulary within 24 hours, and 90 percent within a week. The curve is steeper for Mandarin than for Latin-alphabet languages because each Chinese word is a new shape, not a recombination of letters you already know.

Most vocabulary apps ignore this. They show you a card, you flip it, you press “next,” and the app moves on. The result is exactly what you have lived through: a queue of 500 reviews you cannot sustain, followed by burnout, followed by deleting the app. The 1987 algorithm that most apps shipped until recently, SM-2, was designed by one person for his own learning and never personalised to yours.

The vocabulary apps that work in 2026 do three things differently:

1. They model your actual memory instead of pretending every word is identical.
2. They schedule reviews at the moment you are about to forget, not on a fixed calendar.
3. They source new words from real reading, not from a frequency list you have never seen in context.

Get those three right and the curve flattens. Get any of them wrong and you are back to the queue-and-burnout cycle.

The spaced repetition breakthrough: FSRS vs SM-2

Spaced repetition is the idea that you should review a card at increasing intervals, just before you forget it. The trick is knowing exactly when that “just before” moment is. The two answers that matter for Mandarin learners in 2026 are SM-2 and FSRS.

SM-2 was published in 1987 by Piotr Wozniak as part of his SuperMemo software. It assigns every card an “easiness factor” starting at 2.5. When you press Again, the easiness drops and the interval shrinks. When you press Easy, both grow. It is simple, easy to implement, and good enough that Anki shipped it for almost two decades.

The problem with SM-2 is that it does not actually model memory. It does not differentiate between a card you have reviewed 100 times and one you have reviewed 5. It cannot tell that 的 (de) and 艘 (sōu) behave very differently. It has no concept of difficulty, only of how recently you got the card wrong. And its intervals were calibrated on Wozniak’s own review history in the 1980s, not on yours.

FSRS (Free Spaced Repetition Scheduler) was published by Jarrett Ye in 2022 and is now the recommended algorithm in Anki, Mochi, RemNote, and a handful of other serious flashcard apps. It is free in two senses: the price is zero, and the source code is MIT licensed so anyone can read or fork it.

The headline difference is that FSRS models the forgetting curve as a real function, not as an easiness multiplier. Every card carries three numbers:

• D (Difficulty) - the inherent “hardness” of the card, on a 1 to 10 scale.
• S (Stability) - how many days until recall probability drops to 90 percent.
• R (Retrievability) - the current probability that you recall this card right now.

When you press Again, S collapses and D climbs. When you press Easy, S grows fast. R is recomputed every time you open the app and decays smoothly toward zero as the next review date approaches. The scheduler then picks the next review so R lands near your target (usually 90 percent).

The clever part is the parameter fit. The first few dozen reviews use FSRS’s defaults, which were trained on tens of millions of real reviews shared by Anki users who opted in. After a few weeks, FSRS switches to a model fitted to your history using maximum likelihood estimation. The algorithm literally learns your memory.

The published result, validated against real Anki review logs, is that FSRS users do 20 to 30 percent fewer reviews per day than SM-2 users at the same recall rate. For Mandarin, where a working vocabulary is 3,000 to 5,000 words and you will be reviewing for years, that saving compounds into hundreds of hours.

For a deeper technical walkthrough, see the FSRS explained article.

Why reading is the highest-yield vocabulary activity

Even the best scheduler cannot save a vocabulary built from the wrong input. If you import the HSK 6 frequency list and start reviewing, you will remember maybe 60 percent of the words a year later - and you will not be able to use any of them in a sentence, because you learned them out of context.

This is the deepest lesson of second-language acquisition research: vocabulary sticks when it is anchored to a sentence, a situation, and a story. Words learned in isolation decay faster than words learned in context, even when the review schedule is identical. The 80 percent comprehension band (the “i+1” zone, where you understand 80 to 90 percent of a passage) is where new vocabulary actually moves into long-term memory.

Reading also fixes the most common Mandarin failure mode: knowing a character in isolation but not recognising it in a sentence. Mandarin has roughly 3,500 high-frequency characters, and most of them appear in compounds you cannot guess from the parts. The only way to learn the compounds is to see them in real sentences, repeatedly, across multiple contexts. A frequency list cannot do this. A graded reader can.

The practical implication is that your vocabulary system should look like this:

• Input source. A graded reader (HSK 1-2 texts for beginners, HSK 3-4 folk tales for intermediate). Public-domain stories like 守株待兔 (Waiting for a Hare), 拔苗助长 (Pulling Up Seedlings), and 乌鸦喝水 (The Crow Drinks Water) work well.
• Marking. As you read, tap or long-press every word you do not recognise. Add it to your review queue.
• Review. A modern scheduler (FSRS) shows each marked word at the moment you are about to forget it.
• Audio. Every word you mark should be heard in its tonal pronunciation, because Mandarin is tonal and reading-only vocabulary fails in conversation.

If you can do those four things daily, you will build a vocabulary that sticks. If any of them is missing, the curve bites.

The four pillars of a working vocabulary system

Most “best Chinese flashcard apps” lists compare apps on features. The more useful comparison is on which pillars they cover. A working vocabulary system needs all four.

Pillar 1: Real-world input

Input means reading (or listening to) real Chinese at a level you can mostly understand. Frequency lists do not count. Textbook dialogues do not count, because they are written to teach grammar, not to be read. What counts is text you would read for its own sake if it were in your native language.

Pillar 2: Active marking

Marking means that as you read, you identify words you do not recognise and add them to the review queue. This is the bridge from “passive recognition” (you have seen this word before) to “active vocabulary” (you can use it). Without active marking, the vocabulary in the text never enters your long-term memory.

Pillar 3: Smart scheduling

Scheduling means the review intervals are personalised and adaptive. SM-2 is the legacy baseline. FSRS is the modern best-in-class. Anything that uses a fixed interval (every Monday, every day) is not a scheduler, it is a calendar reminder.

Pillar 4: Daily review

Review means a short daily session where the app shows you the words it predicts you are about to forget. Ten to twenty minutes a day is enough. Skipping days is expensive: FSRS recalculates R from elapsed time, so a one-week gap means more words are “about to be forgotten” the next day, which makes the review queue longer, which makes you more likely to skip again. The daily habit is the load-bearing habit.

A free Chinese vocabulary builder that covers all four pillars is rare. Most cover two or three and quietly expect you to handle the rest yourself.

Free app comparison table

Here is how the leading free vocabulary apps stack up on the four pillars, plus a few practical columns. Prices and free-tier limits change often, so double-check before you commit.

A few notes on the table. ReadDeck is the only app that combines all four pillars out of the box with zero setup. Anki is more powerful but requires an hour of setup before you review a single card, and the daily queue can balloon to 500+ reviews within weeks if you overcommit. Pleco is the best Chinese dictionary on any platform, but its flashcard module is a separate purchase and the review system is less sophisticated than FSRS. Quizlet and Memrise are fine for casual study but their SRS implementations are lightweight compared to Anki or FSRS. Skritter is a niche tool for learners who need handwriting, and it is expensive.

For most learners looking for a free Chinese vocabulary builder, the choice comes down to ReadDeck (zero setup, reading-first) or Anki (more power, more setup). If you want FSRS-quality scheduling without the spreadsheets, ReadDeck is the answer. If you want to study classical Chinese, handwriting, or specialised vocabulary (medicine, law, finance), Anki still wins on flexibility.

The ReadDeck approach: FSRS plus reading plus bottom-curtain reviews

ReadDeck is the app we built because we hit the same walls with every other vocabulary tool. It is offline-first, free, and Mandarin-only by design. The reading experience is the product; the flashcards are a side effect.

How ReadDeck handles the four pillars

Input. ReadDeck ships with three Chinese folk tales (守株待兔, 拔苗助长, 乌鸦喝水) plus an importer for any .txt file you want to read. The stories are short, repetitive, and full of HSK 3-4 vocabulary. For beginners, you can import the HSK 1 and HSK 2 word lists as plain text and turn each list into a reading session.

Marking. As you read, you long-press any word to mark it Unknown. The word is added to your personal review queue, colour-coded blue in the text so you can scan for unrecognised words on a second pass. You can also tap any word to see its pinyin (nǐ hǎo), meaning, and hear its tonal pronunciation without leaving the story.

Scheduling. ReadDeck uses an embedded FSRS 3.0 scheduler. You do not have to configure anything. The algorithm learns your memory from the first review and personalises every interval from there. Because the scheduler is embedded in the binary, no scheduling call leaves your device and no account is required.

Review. The “bottom curtain” review surfaces a small batch of words each morning - typically 20 to 40 cards, sized to fit a 5 to 10 minute session. The words you struggle with come back sooner; the words you know drift to longer intervals. This is what FSRS’s per-card memory model was designed for.

What ReadDeck does not do

ReadDeck is not trying to be Anki. It does not have custom card templates, cloze deletions, or an add-on ecosystem. It does not have thousands of community decks. It does not have handwriting practice. If you need any of those, Anki is still the right tool.

What ReadDeck does is make the daily habit possible. A typical ReadDeck session is 10 minutes reading and 5 minutes FSRS review. That is a commitment most learners can sustain for years, which is what actually builds vocabulary.

For audio-first learners, Audio Deck ships synchronised transcripts that feed into the same review queue. You listen first, read along, mark the words you missed, and review them the next morning. One review queue, two input sources.

For self-testing, the HSK quiz lets you drill vocabulary by level and see which words are sticking and which are leaking.

A 60-day vocabulary plan (HSK 1 to HSK 2)

A good plan is specific. Here is a 60-day plan that takes a true beginner from zero vocabulary to a working HSK 1-2 base (around 300 words and 150 characters) using free tools.

Days 1-14: Foundation

• Daily commitment. 10 minutes of reading + 10 minutes of FSRS review.
• Reading material. Import the HSK 1 word list as a .txt file in ReadDeck, or use the first two pages of any beginner textbook. Each sentence should be 80 percent comprehensible.
• What to mark. Every word you cannot pronounce or do not recognise. Do not try to be selective. Over-marking is fine; under-marking is the bigger risk.
• Review. Every morning, do the full FSRS queue. The queue will be small (10-20 cards) for the first week and grow to 50-80 cards by day 14.
• Audio. Every time you mark a word, listen to its pronunciation in ReadDeck or via the Audio Deck beginner track. Mandarin is tonal; reading without audio builds a vocabulary that fails in conversation.

Days 15-30: Compounding

• Daily commitment. 15 minutes of reading + 10 minutes of FSRS review.
• Reading material. Add the first folk tale (守株待兔) to your ReadDeck library. Read it once for comprehension, then again to catch the words you missed.
• What to mark. Continue marking everything you do not recognise. By day 30 you should have 150 to 200 marked words.
• Review. Queue size will plateau around 100-150 cards. This is normal. Do not panic; trust FSRS.
• Self-test. Take the HSK 1 quiz once a week to confirm retention. If you score below 80 percent, slow down the new-word rate, not the review rate.

Days 31-45: Second input source

• Daily commitment. 15 minutes of reading + 10 minutes of FSRS review.
• Reading material. Add the second folk tale (拔苗助长) and start an Audio Deck beginner track.
• What to mark. Add words you missed while listening. If you can recognise a word in audio but not in text, that is a useful data point - it tells you where to focus.
• Review. Queue size will briefly spike (to 200+ cards) as the new words enter, then settle. This is FSRS catching up. Do not skip days to “reset” the queue; that breaks the algorithm.

Days 46-60: HSK 2 territory

• Daily commitment. 20 minutes of reading + 10 minutes of FSRS review.
• Reading material. Add the third folk tale (乌鸦喝水) and import the HSK 2 word list as a .txt file.
• What to mark. Everything you do not recognise. By now you should be marking fewer HSK 1 words (they are entering long-term memory) and more HSK 2 words.
• Review. Queue size will grow to 250-350 cards. This is the inflection point where most learners quit. Do not. The compounding effect shows up between day 60 and day 90.
• Self-test. Take the HSK 2 quiz on day 60. A score of 70 percent or higher means the system is working.

By the end of day 60, you should be able to read a short HSK 2 passage with maybe 5 to 10 dictionary lookups per page, recognise 250 to 300 words in context, and pronounce most of them correctly with the right tone. That is enough to start HSK 3 content with momentum rather than dread.

Common pitfalls

Even with the right tools, most learners slip on the same handful of habits. Watch for these.

Skipping audio

Mandarin is a tonal language. Reading without listening builds a visual vocabulary that fails in conversation. Pair every reading session with audio. In ReadDeck, every word has tap-to-play audio; in Audio Deck, every lesson is narrated with adjustable speed.

Adding too many new words per day

The temptation is to import the full HSK 4 list and “get through it faster.” FSRS will schedule all of them, the queue will balloon to 500+ cards, and you will burn out within two weeks. The sustainable rate is 10 to 20 new words per day for beginners, dropping to 5 to 10 as the deck grows.

Marking nothing

The opposite failure mode is reading without marking. If you only review words from a pre-made frequency list, you never anchor them in context, and retention collapses within a month. Mark every word you do not recognise, every time.

Using SM-2 in 2026

If you are on Anki and have never switched to FSRS, switch. The migration takes five minutes: open Deck Options, change the algorithm to FSRS, click “Optimise” once you have 1,000 review logs, and pick a target retention between 0.85 and 0.90. You will see interval changes on day one. This is normal. Trust the algorithm.

Treating the queue as the enemy

The daily review queue is not a punishment. It is the system working. If you skip days to make the queue smaller, the queue grows back faster the next time you open the app (FSRS recalculates from elapsed time). The discipline that works is the boring one: do the queue every day, even when it is 100 cards, even when you are tired.

FAQ

What is the best free Chinese vocabulary builder?

For most learners in 2026, the best free Chinese vocabulary builder is the one you will actually use every day. ReadDeck is the strongest reading-first option because it requires zero setup and embeds FSRS by default. Anki is more powerful but requires an hour of configuration before you review a single card. Choose the one that matches your tolerance for setup.

Can I build a Chinese vocabulary without Anki?

Yes. Anki is a general-purpose flashcard engine; it does not source vocabulary from anywhere in particular. You can build a working vocabulary by reading graded texts, marking words you do not recognise, and reviewing them with any modern scheduler (FSRS in ReadDeck, FSRS in Anki, or Pleco’s SRS). The tool matters less than the daily habit.

How long does it take to build a 1,000-word Mandarin vocabulary?

With a working system, 3 to 4 months. The HSK 1-3 lists together cover about 1,000 words. At a sustainable rate of 10 to 20 new words per day, plus daily review, you will hit 1,000 marked words within 90 to 120 days. The mistake is trying to go faster. Going faster produces a queue you cannot sustain.

Do I need to learn characters and words together?

Yes. In Mandarin, the character is the word for the vast majority of the HSK 1-3 vocabulary. Learning words as pinyin-only is a trap: you will remember the sound but not the character, and you will not be able to read. The flashcard should show the character as the primary cue, with pinyin (nǐ hǎo) as a support, not the other way around.

Is Pleco enough on its own?

Pleco is the best Chinese dictionary ever made for a phone and should be in every learner’s toolkit. It is not, by itself, a vocabulary builder. The flashcard module is a paid add-on, the SRS is less sophisticated than FSRS, and there is no graded-reader pipeline. Most advanced learners pair Pleco with a dedicated reader app (ReadDeck, Du Chinese, or LingQ).

How does FSRS compare to Anki’s old SM-2 algorithm?

FSRS is a strict upgrade. For the same target recall rate, FSRS users do 20 to 30 percent fewer daily reviews than SM-2 users. The reason is that FSRS models the forgetting curve as a real function (difficulty, stability, retrievability) rather than as a single easiness multiplier. For Mandarin, where the deck is large and the word difficulty varies widely, the savings are largest.

Where to go from here

If you read this far, you already understand the system: read real Mandarin, mark what you do not recognise, let FSRS schedule the reviews, do the daily queue. The only thing left is to pick a tool and start.

ReadDeck is the free, offline, FSRS-powered reader we built for this exact workflow. Install it, open one of the folk tales, long-press a word you do not recognise, and let the scheduler do its work. The first session takes about 15 minutes. By session ten, the habit is forming. By session thirty, the compounding effect is visible.

For audio-first study, pair ReadDeck with Audio Deck. For self-testing, drill your level on the HSK quiz. For a deeper look at the scheduler powering the system, read what FSRS is and how it differs from SM-2.

The vocabulary that actually sticks is the one you build one word at a time, in context, with a scheduler that respects your memory. The tools are free. The only thing that costs is the daily habit.

For related guides, see the Mandarin vocabulary hub, the Anki vs ReadDeck comparison, and the Chinese reading practice for beginners guide.