[koreth1]

This looks really useful! Wish I'd had something like this when I was learning Mandarin.

I'm curious what determines whether or not you add a given language to the list. DeepL and Claude, at least, have usable translation ability in more languages than the app currently supports. Is there a lot of manual effort required for each language, or do you want to keep the list limited just to avoid overwhelming users?

[Alex-Programs]

Thanks! I'm glad people like it; I'm hopeless at marketing it to Language Learners:tm:, but programmers seem to love it and it's nice to get some positive feedback.

DeepL is actually pretty limited in what it supports. Unless I've missed a new language, Nuenki supports all of DeepL's languages.

Some of the additional ones are supported via Claude only and, where permitting, Groq. Groq is far faster than Claude; in languages that DeepL supports, DeepL handles visible text and Claude handles text that you haven't scrolled to yet. Claude-only languages are a bit of a worse experience.

It's pretty easy for me to add a language. It's all stored in a centralised toml file, which happens to be open source - https://github.com/Alex-Programs/nuenki-languages/blob/maste... - and it's about a 20 minute job to add a language, test it, etc. Then it's about half an hour and 5 USD to benchmark whether Llama is any good at it, and if so enable Groq and make the experience a bit more pleasant. I'm currently working on improving the translation quality benchmark (https://nuenki.app/blog/the_best_translator_is_a_hybrid_tran...), because people seem to like it and there's definitely a lot of room for improvement.

That 20 minute number is without updating the big language cloud on the website, which is a bit finicky; iirc I haven't added Vietnamese to it yet.

If anyone here has any requests, I'd gladly add them!

[foolinaround]

A request to add Tamil, a widely spoken and one of the earliest classical languages! Thanks!!