A vocabulary builder app helps you learn and retain words in a foreign language. That description covers tools that work in genuinely different ways, and the differences matter more than any feature list, because they determine which words you end up knowing. This article sorts the category into four types, says who each one is actually for, and is explicit about where the one I build fits and where it does not.
The two questions that decide everything#
Strip the marketing away and every vocabulary tool is answering two questions.
Where do the words come from? A prebuilt deck, a frequency list, a corpus, or the things you personally read. This decides relevance. A word you met in an article you chose arrives with a reason to remember it. A word from a generic top-2000 list arrives without one.
How do they come back? A schedule, a game, a re-read, or nothing at all. This decides retention. Hermann Ebbinghaus mapped the forgetting curve in 1885 and the shape has held for 140 years: without scheduled repetition, most of what you learn today is gone within a week.
Almost every tool in this category is strong at one and weak at the other. Anki has the best-studied scheduler in the field and supplies no words at all. A browser extension captures words beautifully at the moment of curiosity and does nothing to bring them back tomorrow. Recognising which half a tool solves tells you what you need to pair it with.
Type 1: curated decks and courses#
Memrise, Drops, Vocabulary.com, Quizlet, most textbook companion apps.
Someone has already chosen the words, ordered them, and often illustrated them. You open the app and start. Drops leans on imagery and a five-minute timer, Memrise on video clips of native speakers, Vocabulary.com on English test prep with unusually good definitions.
The strength is the zero-decision start. On day one you do not know enough to choose your own words, and being handed a sensible first 500 is worth a great deal.
The ceiling arrives sooner than people expect. Curated decks tend to top out in the low thousands of words, and the top few thousand of any language are the easy part — they are frequent, so you meet them constantly and they stick almost by accident. The words that actually gate reading are the mid-frequency ones, and they are different for a reader of crime novels than for someone reading football coverage. No curated deck can be relevant to both.
Best for: absolute beginners, and anyone who wants a structured path more than a personalised one.
Type 2: corpus sentence apps#
Clozemaster, Lingvist, Speakly.
These pull sentences from a large corpus, order them by frequency, and drill you on one word per sentence with the rest as context. Clozemaster does this at enormous scale and is close to free; Lingvist adds an adaptive model that guesses what you already know.
The design is smart. A word inside a sentence gives you collocation and register alongside meaning, which a bare word pair cannot. Frequency ordering means you are always working on the most statistically valuable word available.
The complaint that comes up consistently is that the sentences are disconnected. You are working through a stream of unrelated statements about unrelated things, which is efficient and, for many people, dull enough to end the habit. There is also no reason to care about any individual sentence, and caring is what memory runs on.
Best for: intermediate learners who want frequency-ordered coverage and do not mind drill.
Type 3: capture tools#
Browser extensions, share sheets, popup dictionaries like Yomitan, OCR scanners.
These do not supply words. They make it nearly free to grab one at the instant you meet it — select and right-click on a web page, share from any app on your phone, hover for a dictionary entry, photograph a menu.
The timing advantage is real and underrated. A word you look up in the sentence that made you curious arrives with meaning, register and grammar attached. The same word typed into a deck three hours later arrives naked, and you have already forgotten why you wanted it.
The structural gap is review. Almost nothing in a browser tab schedules a word to come back tomorrow. Capture tools need to be paired with something that does, and the pairing is where most setups fall apart — words end up in a list nobody opens.
Best for: everyone, as the second half of a pair. Rarely sufficient alone.
Type 4: reading-first apps#
LingQ, Readlang, Beelinguapp, LingoBlend.
Here the source is your reading. You read a text, the unknown words are made cheap to resolve, and the ones you save become your deck. The list is personalised by construction, because it is a record of what you actually failed to understand.
LingQ imports full texts and tracks a cumulative known-word count across a large community library. Readlang layers hover-translation over the web pages you already read. Both assume you can already parse most of a target-language page, which is a fair assumption from roughly B1 and a punishing one below it.
LingoBlend takes the opposite on-ramp. Rather than asking you to read a full target-language text, it takes text you already understand and swaps a percentage of the words — you set it with a slider — into the language you are learning. You read a sentence that is mostly comprehensible with a few target words woven in, tap one to see its meaning and grammatical form, and save it. As reading gets easier you push the slider up. This is the diglot weave, described by Robbins Burling in 1968, and the mechanism is covered in what is the diglot weave method.
The same thing happens in the browser. The extension blends live web pages on desktop Chrome and Edge with the same 5-to-100% slider, translating on-device through Chrome's built-in translator, so page text never leaves your machine. Words you click and save land in the same dictionary as everything else.
For advanced readers there is a library of unabridged classics — Don Quijote, Les Misérables, Pride and Prejudice, Dom Casmurro, Der Prozess, Pinocchio and others from Project Gutenberg — read chapter by chapter with every word tappable. More on that in reading classic literature in another language.
Saved words, wherever they came from, run through five games on an Anki-style SM-2 schedule with sub-day learning steps. Where it is weak: if you want a large community library of shared lessons with audio, LingQ has a decade of head start and this is not close.
Best for: learners who already read something in their own language and would rather convert that time than add a new study block.
The four types, side by side#
| Curated decks | Corpus sentences | Capture tools | Reading-first | |
|---|---|---|---|---|
| Word source | Editor's list | Frequency corpus | Whatever you meet | What you chose to read |
| Personalised | No | By frequency only | Fully | Fully |
| Context per word | Usually none | One sentence | The real sentence | The real text |
| Built-in review | Yes | Yes | Usually none | Yes |
| Works at A0–A1 | Yes | Poorly | Poorly | Varies |
| Realistic ceiling | ~2,000 words | Large | Unlimited | Unlimited |
| Effort to start | None | None | Low | Low |
| Typical failure | Words you never meet again | Boredom | Words never reviewed | Requires a reading habit |
How many words are you aiming at#
This is the number that makes the deck ceiling concrete.
Paul Nation's coverage research asks how much of a page you need to know before reading stops being decipherment. The answer is around 98% of running words for comfortable unassisted reading, which corresponds to roughly 8,000–9,000 word families — with about 3,000 covering 95% of everyday text, enough to follow the gist with frequent lookups.
Set that against a curated deck of 2,000 words and the arithmetic answers the question for you. Decks are an on-ramp, not a route. Somewhere in the low thousands, the only remaining source of relevant words is the text in front of you, which is why every advanced learner eventually converges on some version of read-and-capture regardless of where they started. How many words to be fluent in Spanish works the numbers through for one language, and measure your vocabulary size covers how to find out where you currently stand.
Choosing, in three questions#
Can you read a page in your target language and understand most of it? If no, start with a curated deck or an adjustable blend. If yes, a reading-first tool will outperform any deck.
Do you already read something regularly? News, fiction, forums, documentation. If you do, a tool that converts that existing time beats one that asks for a new half hour. If you do not, a scheduled deck gives you the structure you are missing.
What happens to a word after you save it? If the honest answer is nothing, fix that before changing anything else. Capture without review is the most common failure in this whole category, and it is invisible until you notice your saved list has 400 words and you recognise 40 of them. Spaced repetition for language learning covers what a schedule needs to do.
Feature sets and prices change on every product named here. Check each one's own site before deciding; LingoBlend's tiers are on the pricing page.