Vocabulary, out of your own life

Catch words
from the wild.

Hear a word you don't know. Say it out loud and keep going — it comes back as a card set in the moment you caught it: the film, the podcast, the meeting where it first went past you.

About a minute to set up · No account needed to try

Catchword's Capture screen in the iPhone simulator: the active session at the top and a large microphone button reading 'Tap to speak'
Actual app · iPhone 17 Pro simulator
~15
unknown words in a single film
<5s
to catch one without stopping
5
spaced reviews before it sticks

The problem

Your words, not a list someone else wrote.

The words worth learning already found you — in a film, on a call, in something a colleague said twice before you admitted you didn't know it. Looking one up costs you the moment. Not looking it up costs you the word. Most people pick the moment, every time, and that is the right call — it just leaves nothing behind.

One ordinary day

Words you meet and don't know
~15
Words you stop to look up
1–2
Words you still know next week
0

How it works

Four seconds, then nothing.

No deck to build, no cards to write, no evening set aside for review. The work is a word said out loud and two minutes a day.

  1. 01

    Keep doing what you're doing

    Name a session for whatever you're in — a film, a podcast, a stand-up, a call. Catchword sits in the background and does nothing until you need it.

  2. 02

    Say the word

    A word goes past that you don't know. Say it out loud. That's the entire interaction — under five seconds, without leaving the moment you're in.

  3. 03

    The card writes itself

    Translation, part of speech, a plain definition, and examples set in whatever you were doing when you caught it. You didn't type anything.

  4. 04

    It comes back five times

    One day, three days, a week, two weeks, a month. You answer before you're shown. That order is what moves it into memory.

Catchword's Progress screen in the iPhone simulator: 55 words captured, 55 due today, 100% recall, and a bar per spacing interval
Actual app · iPhone 17 Pro simulator

Capture

Your voice is the whole interface.

Say the word. No pausing, no typing, no asking anyone to repeat it. The one thing standing between hearing a word and keeping it is effort, so we removed almost all of it. Speak it in any of sixty languages, and read it back in any of the other fifty-nine.

Works on words you can't spell — which, for a word you've only ever heard, is most of them.

Catchword's Capture screen in the iPhone simulator: the active session at the top and a large microphone button reading 'Tap to speak'
Actual app · iPhone 17 Pro simulator

Context

Every word remembers where it came from.

Cards are set in whatever you were doing when you caught the word, not in a generic sentence from a textbook. You don't recall 'relentless' — you recall the wave in the film that never let up.

Provenance is most of why a word sticks. Stripping it out is the quiet mistake every word list makes.

Catchword's Sessions screen in the iPhone simulator, listing a session per film — Dune: Part Two, Oppenheimer, Anatomy of a Fall, The Bear, Chernobyl, Past Lives — each with the date watched and the number of words caught
Actual app · iPhone 17 Pro simulator

Recall

Answer first. Then find out.

Every review asks before it shows. Retrieving a word — even failing to — is what moves it into long-term memory; rereading it does close to nothing.

One day, three days, a week, two weeks, a month. You never choose an interval; forgetting is predictable enough that we can schedule against it.

Catchword's Review screen in the iPhone simulator: 55 words due today above the one-day to one-month spacing schedule
Actual app · iPhone 17 Pro simulator

The method

Forgetting is predictable.

We schedule against it.

Built on the boring, replicated stuff.

None of this is novel. It is the part of the research that survived replication, implemented properly instead of gamified into streaks.

Successive relearning

Not one study session, but the same word retrieved correctly across several sessions spread over weeks. It is among the most reliably replicated findings in learning research, and almost nothing consumer-facing implements it properly.

Retrieval before reveal

Being shown a word teaches you that you recognise it. Being asked for it teaches you whether you know it. Only the second one builds recall, so every review is a question first.

Immediate feedback

The answer arrives the moment you commit to one. A wrong guess corrected within a second is worth more than a right one you were never sure about.

Pricing

One word free. Then all of them.

Catch a word, see the card it becomes, and decide from there. The free card is the whole product in miniature — there is no trick to it.

What you getFreePro
Words you can catch1
Instant AI cards1
Examples set in context
Spaced review schedule
Synced across devices
Start catching words

Build your plan first — you'll see the price before anything is asked of you.

Questions

The things people ask.

Is this only for movies?

It started that way, and an evening of film is still the densest source there is — which is why most of what you see here is filmed in one. But a session is only a name: a podcast, a stand-up, a call with a colleague, a chapter you're halfway through. You catch a word the same way wherever you are, and the card is set in whatever you called it.

How is this different from Anki?

Anki is a scheduling engine that assumes you'll make the cards. Most people stop making them within a fortnight — that's the failure mode, not laziness. Catchword removes card-making entirely: you say a word and the card exists. The scheduling underneath is the same well-tested idea.

Do I have to stop what I'm doing?

No. Saying a word takes under five seconds and you don't step out of the moment to do it. The card is generated while you carry on, and it's waiting for you afterwards.

Which languages does it support?

Around sixty, in any combination. You pick the language you're learning and the language you'd like to read translations in — English for a Brazilian, Japanese for a Spaniard, whichever way round you need. Cards are written for the pair you chose rather than pulled from a fixed dictionary, so no combination is second-class.

What if it mishears me?

You can correct a word before the card is generated. Accents and background audio being what they are, this happens — it's a two-second fix, not a lost word.

Does it work offline?

Reviews do — your cards are on the device. Catching a new word needs a connection, since the card is generated server-side.

What happens to my data?

Your words and where you caught them are yours; we don't sell them or train public models on them. Audio is transcribed and discarded rather than stored. The full detail is in the privacy policy.

The next word you don't know is about ninety seconds away.

Build your plan, catch one word, and see what the card looks like. It takes about a minute and costs nothing.