Why I made LingoChunk (and keep adding to it)

English isn’t my first language. I grew up with Ukrainian and Russian, and I’ve lived in the UK for over twenty years, so my English is fine. A few years ago I got curious about doing it the other way round: picking up a language from scratch as an adult, trying different techniques, and seeing what actually stuck for me.

I went with Greek. I love the culture and I keep going back to Greece, and on the usual “how hard is this language” charts it sits somewhere in the middle, so it felt like a sensible place to start. That was about three years ago.

It went fine for a while. I tried a bit of everything: courses, apps, audio, things I found on YouTube. I’ve always been an Anki person, so whatever I was studying with, I’d end up turning it into cards. And then I’d run dry. Some of that is just how motivation goes, but a lot of it is Greek specifically: there isn’t much material compared to the big languages. My wife is learning German, and the amount of stuff she has to play with is on another level. I finished the entire Greek tree on Duolingo and it felt thin next to what German learners get, and it was the same story with most of the input I went looking for.

So about six months ago I switched to German, mostly to see what this all feels like with the firehose of material turned on. I’m still learning by ear, still making cards, just with a lot more to chew on.

None of that is really why I’m writing, though. The point is the one thing I kept reaching for the whole time, in both languages: a decent way to keep what I hear.

I’m a flashcard person. When I hear something worth keeping (a word, the way it’s said, the sentence it sits in) I want it in a deck, with that exact audio. And pulling that out of a recording by hand is close to hopeless. You can’t select a word you heard. You can’t click a sentence. It’s just sound going past.

So I started messing about with ideas to fix that, for myself.

It started as a flashcard hack

At first I only wanted one thing: grab a word or a phrase I cared about, with the audio it came in, and turn it into a card. Nothing clever.

Then I got into shadowing. The short version is that you repeat what you hear out loud, right behind the speaker, copying the rhythm and the sounds. The people who can explain it properly already have, and it usually gets traced back to the polyglot Alexander Arguelles, so I won’t pretend to be your teacher here. I’ll just tell you what I needed.

The way I shadow, you start from the end. You take the last little piece of a sentence, the smallest chunk you can say cleanly, and you repeat it until it’s comfortable. Then you grow it backwards until you can say the whole thing without tripping. Not word by word, that’s too granular. You add the next natural chunk, as much as still feels comfortable to say in one go. You build from the end, not the start.

And this only works if it’s instant. Tap the end of the sentence, loop it, then extend the loop a chunk at a time, with no scrubbing along a timeline and no counting seconds. If setting up each loop meant rewinding and fiddling for ten or twenty seconds, I’d never do it. I know that, because that’s exactly what made me give up before.

I went looking for an app that let me do this: loop a sub-second piece, grab it from the end, expand it backwards, all instantly. I never found one that did it the way I wanted. So, being a developer, I built it.

The actual idea: almost no friction

This is the part I care about most, and it isn’t really a feature. It’s a rule I hold the whole app to. The gap between wanting something and having it should be a second or two. No copying. No hunting around. No “now go and open this other thing first.”

Because I can select anything in the transcript, down to a single word, that one idea opens a lot of doors. I select a sentence and ask for the grammar, and two seconds later I’m reading a proper, formatted explanation. I select a phrase and ask for more examples like it. I select a word and it’s a card. Same gesture, different outcomes, all instant.

That speed quietly changes how much I do. When a lookup costs two seconds instead of two minutes, I look things up all the time, instead of promising myself I’ll do it later and then never doing it.

What I actually do every day

Two habits, mostly.

Normally I go from the start of the audio and work forward. When I hit a word I don’t know, one click and it’s a card, with the real audio examples taken from that same audio. One or two seconds, every time, and I’m back to listening.

Other days I don’t listen first. Once the app has processed an episode it hands me the full list of words it found, in their dictionary form. I scroll through and click the ones I don’t know, one by one. Each click is a card. It’s oddly satisfying.

There’s also a way to grab a whole set by level, “give me the B1 words from this episode”, in a single click. Honestly I don’t reach for it much, but it’s there.

If you’d rather see the whole thing in motion than read about it, here’s a short walkthrough:

Fair warning: it’s already a little out of date. The app moves faster than I can re-record, so a few things will look different by now. The basic ideas are all still there, though.

Why I started building my own reviews

At first everything went straight to Anki, and that still works. The export carries the full audio examples, so if you live in Anki, nothing’s lost.

But there’s one Anki limitation that bothers me more than any other. Once a card is made, it’s cut off from the audio it came from. The card sits on its own, with no way back. I wanted the opposite. While I’m reviewing an example, I want one click to see it again in the transcript, in its wider context, the whole passage around it and not just the single line.

And I wanted the things I can already do while listening to be there on the card too: explain the grammar, give me more nuance on a word, the rest of what modern AI is good at. Some of that you could probably bolt onto Anki with plugins. But I didn’t want it bolted on. I wanted it in one place, integrated with the transcript it came from.

So I built a small spaced-repetition system inside the app. It’s good enough now that I do almost all of my new studying in it. The Anki export is still there if you’d rather live in Anki.

The newest thing I’m playing with

Lately, when I ask for a grammar explanation, the app also offers cards for that exact concept. One click and they’re in my deck, ready to review, and exportable to Anki like everything else. So a thing I didn’t understand a moment ago becomes something I’m studying, in a click, while I’m still looking at the explanation.

Here’s the half-minute version: select a phrase, ask, and read the grammar right there in the app, no jumping out to ChatGPT.

It’s not polished yet. I use it anyway.

The languages I can’t vouch for

The app handles a long list of languages, and I speak almost none of them, Chinese and Japanese included. I can only really vouch for the pairs I use myself; for the rest I’m trusting automated tests, not my own eyes. So if you study one of those and something looks off, I’d love to hear about it.

Still building, and sharing it

None of this is finished. It changes most weeks, usually because something annoyed me during my own studying that afternoon. It stopped being a hack a while ago, though. It’s now the main way I work through comprehensible input in German.

I built it for myself, but I’m putting it out here in case it’s useful to someone else who learns by ear. If you want to poke at it, you don’t need an account to try it on the audio that’s already there. Make a free account and you can upload your own audio and run it through the whole thing. It’s all free while it’s in beta.