Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

(Shameless plug at the end)

During my Erasmus in Germany, I've tried almost all of the top language learning apps (Duolingo, Babel, Seedlang, Anki...) and none of them have really worked me.

What I wanted was:

* Learning in context --> A lot of German words do not have direct English/Czech translation, so learning the 1:1 word translation did not work well

* Having audio for each card

* Intensive pace --> Going through a lot of cards in one session. Duolingo is the worst in this as you spend a lot of time doing super easy childish exercises. If I want to learn let's say 30 min a day, I want to pack as much content as I can into it.

* Skipping ahead --> My level is around B1/B2, I don't want to do placement tests and then re-learning the words I already know

* Learning from my content --> I like to consume German podcast/youtube videos/websites and in some apps, it was quite difficult or impossible to add words I've just encountered "in the wild".

Anki has worked the best, but generating cards with sentences and audio was quite cumbersome and (a seemingly minor detail) I was loosing flow while thinking whether I knew the card "well", "good", "easy" or whatever the options are. What I like better is a simple knew/didnt know, while still using spaced-repetition.

So for the past year, I'm on-off working on a language learning app, currently only supports German, which helps you extract words from content (youtube, web, text), handles different word forms, and then generates cards with infinitely many sentence examples through GPT4, with a nice audio (latest GCP model).

The project website with Android and iOS links: https://vokabeln.io/ (web design is old, app looks very different now)



If you are really overwhelmed by the answers, then you could stick to just one or two of them. I agree that having too many can be overwhelming. The SRS algorithm should still work. I don't recommend to use "well", as you won't see the cards again for a very long time.


I’d recommend using GPT4 to generate English sentences and then Google Translate to get the German. Harder to get example sentences for a specific German word, but may improve the quality of the German. (I had this issue when generating specialized language learning podcasts in Swedish and Polish. Lots of English-isms in GPT4’s output).


I can recommend this too, although I use DeepL instead of Google Translate, and I also use DeepL Write to identify issues in my grammar


Your landing page shows this example:

Der Fußballspieler schießt auf das _____.

Should be “Tor”. Nobody would say “Ziel”.


Also if you are going to give this a try, two features that are missing and frequently requested: - articles and plurals for nouns - wordlist with A1, A2, B1 words for people doing certificates I'm hoping to implement them next month.


> What I like better is a simple knew/didnt know, while still using spaced-repetition.

Could you just click on the two ends, 1min Vs the scaled max time? I.e. ignore the two middle buttons.


It doesn’t cover all of your requirements but you might find Clozemaster interesting


Yes, I've tried that one as well, but text input is too difficult and slow, whereas multiple choice is too easy. A lot of the time, I would be able to click correct option elimination method.


Nice, it would be useful to have this in other languages.


I'll add more, once I figure out how to do at least one language well, but will do in a few months.


Dude. You rock at UX design. This is amazing…

Only trouble I have tbh is figuring out what I want to read in German! But I’ve signed up and will give this a spin, and send what feedback I can.


Thanks! My design is the one on the website. The neo-brutalism was done by a junior designer who was looking for her first project and she did extremely good job!


The real delight point for me was how painless it was to mark 700 words as already known and pickup a bunch of fairly common words i am not comfortable using (like “allerdings” and “eigentlich”) using that scroll-to-set-learned control.


That one took 3 completely different iterations and bunch of tweaking to get right. For example, when you open in, it scrolls a bit to the first word. When it did not scroll, a friend of mine was looking at it helplessly, trying to scroll on the left side with no success. UX is so difficult.


FWIW the one on the website is much cleaner, self-consistent and overall more pleasant to look at it. The one in the app (at least the iOS one) is a clear step down. Just 2c.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: