Show HN: I built this to talk Danish to my girlfriend – works with any language (menerdu.vercel.app)

204 points by lil_csom ↗ HN
I'm in my 4th year living in Denmark as an expat, and I finally decided it’s time to properly learn Danish. I do have a Danish girlfriend, after all. One way I’ve been practicing is by trying to text only in Danish, but I often find myself stuck. I start my message in Danish, then hit a wall because I don’t know a word or how to fit something naturally into the sentence.

Especially in those cases, I used to give up and translate the entire message from English, which kind of defeats the purpose and interrupts the learning process.

So I started prompting GPT. I’d write my message with wildcards or notes for the parts I didn’t know, and it would return a corrected version. That worked well, but reusing the prompt each time became tedious.

So I built a wrapper around it.

Now I can type in the target language, mark unclear parts with curly braces {like this}, and get an instant corrected version with explanations. I also added a history feature so I can review what I got wrong, and I plan to build more on that soon (eg. summary of areas or words to review).

This app is for language learners who want to practice writing without feeling insecure about mistakes or breaking their flow by switching to a translator.

I hope you find it useful!

56 comments

[ 3.3 ms ] story [ 70.6 ms ] thread
Also, yes the TTS doesn't work yet - I decided to share the MVP first before adding my Google billing account :P
I got an error. Is it working now?
First of all, 10 points for the dedication and perseverance! The website looks super professional and promising. What other languages can it help with? I am planning to learn Arabic. Also, are there options to choose between different dialects of a language?
This is really cool!

I've done a lot of research into LLM translation for my product[0], and I'm currently working on a deep translation service that provides reliably human-level translations.

I don't know what model you're using, but GPT-4.1 is probably the best for your use case - it's in the top few % for nearly every language, and has a low standard deviation, while also being relatively low latency and low cost.

[0] https://nuenki.app

In the example text hjelmpe is not a word in Danish.
Tried it with some Dutch, only getting "Unable to process correction" errors. Good idea but seems it need a bit more error handling.
Well I broke it immediately, haha. Sorry about that. Apparently it doesn't like Chinese.
Same, ha. I think he ran out of API credits actually, in my case at least.
This is an interesting idea.

But, besides not being able to use your site due to the errors mentioned by other posters, both the examples on your website give me pause:

The example input "Kan you hjelmpe mig {yesterday}?" reads to me as "Can you help me yesterday?", but that's just a nonsense sentence and an odd choice for an example. The word for "help" is also misspelled, but presumably that would get corrected.

And the suggestion of "Jeg vil gerne handle i morgen" for "Jeg vil gerne go shopping i morgen", instead reads to me as "I would like to act tomorrow". A more idiomatic translation would be "Jeg vil gerne købe ind i morgen".

> And the suggestion of "Jeg vil gerne handle i morgen" for "Jeg vil gerne go shopping i morgen", instead reads to me as "I would like to act tomorrow". A more idiomatic translation would be "Jeg vil gerne købe ind i morgen".

It's a little better, but I would never expect anyone to translate "shopping" to "købe ind". "købe ind" is about getting groceries for the week, shopping is about walking the strip and dreaming of buying random clothing. As a native speaker I'd be less surprised if the you just used the borrowed word "shopping" directly. Basically "Jeg vil gerne shoppe i morgen".

It seems that my GPT credits have been deplated, thus the error :) Sorry about that! I decided that if I do ran out, I will continue the development - showing that is has popularity!

In the meantime, I am at work, so I don't think I can add the rate limiting now... but thank you for everyone for the input!

it would presumably require quite a large rework of what you have, but you could imagine a version of this where the client was talking to the APIs, so that people could input their own API keys without sending them anywhere.

the core idea here is really cool, and the UX is (or seems like it would be) impressive - no need to select a language, for instance.

Good luck getting it back on its feet!

Really good idea.

I also really like the name.

Maybe add a language selector such that the AI has a better context of what language the user is interested in.

i tried a simple "donde (bus station)?" but i got an error about an invalid text
I entered the text that was present in the placeholder, but I got an error message: "Unable to process correction Error processing your text. Please try again."

Quick feedback, the idea sounds interesting to me and I wanted to learn more. Feature request: more examples, or cached examples that would still work if the server crashes.

Also got an error :/ but excited to lean on this as a supplement to the Hindi conversational lessons I've been taking!
I’m sure she is a lovely person, but I don’t really have a need to text your girlfriend.
Sounds cool!

(Why doesn't ChatGPT make something like this, essentially plugins, for their web interface? You make a script with macros and put that on a "Marketplace" and people can run it... would remove the need for setting up a whole nother website just to wrap around prompts)

Great idea and very well executed.

... Now for the hard challenge - try and speak Danish :)

Why is the suggested input misspelled? Is it because it is supposed to help with misspellings?
This is a really cool concept :) Generally in language learning it's a good idea to try to "think" in the language as much as you can so you get creative about how to phrase something with the words you do have, but if that fails, particularly when you're just starting out this is a great motivational aid and conversation starter if nothing else :) I would definitely recommend integrating with an established translator like DeepL as well as a kind of second correction as AI does get things wrong and having the two versions can help compare what may have gone wrong, you can probably keep your autodetecting language bit as well as I am pretty sure DeepL supports language identification and if not, I know Google does. Good luck with learning Danish, I tried my hand at it a few years back but have switched to Finnish as of early this year
Just getting an error: Unable to process correction
I am wondering if it will work for my girlfriend. She speaks English.
A long time ago, before SEO spam, I would Google parts of what I had written phrase by phrase (with proper nouns removed or generalized) and if the number of results was under a certain threshold, I'd assume I was wrong.
I really hate the term “expat”
(comment deleted)
It doesn't seem to work with "any language", as claimed. It cannot process Chinese writing (nor can it process pinyin). "Error processing your text", is the message. Though it's certainly possible the volume of traffic has overwhelmed things and it would work under normal circumstances?