Open okjodom opened 8 months ago
I hackily repurposed a script for a different project to do https://github.com/fedimint/ui/pull/205, I'll try to post that here so that someone (maybe me if I find time) can make it re-usable.
I've been exploring ways to formalize and automate the Fedimint UI project's translation pipeline to streamline the localization process. After some research and analysis, I'd like to propose a solution inspired by successful implementations in projects like Blink Wallet, Bitcoin Core and Jam UI.
Justification
Implementation Plan
I have developed a quick POC to validate the proposed localization pipeline using Transifex and i18next, which involved setting up a React project, as well as integrating Transifex and i18next.
@Kodylow
What's wrong with doing the translation scripts, LLM translation got a thousand times better in the last couple months and costs like 30 cents to do a full translation? I updated it a little while ago and with a Google API key it takes like 5 mins. Is transfinex just for reviewing those or for doing the full translations?
Transifex allows both reviews and full translation - emphasizes a collaborative approach.
What's wrong with doing the translation scripts, LLM translation got a thousand times better in the last couple months and costs like 30 cents to do a full translation? I updated it a little while ago and with a Google API key it takes like 5 mins. Is transfinex just for reviewing those or for doing the full translations?
Okay that seems like a big lift for minor gain, there's no major prose/educational content that's specific to get correct, almost all of it is word or phrase level translation that we can like 99% of the way through with translation scripts. If you make the pipeline it might be useful in the future if we add more translatable content but generally seems unnecessary right now.
we need to standardise how we generate translations from
en
to other supported languages. ideally this is an automated process that fits into the current i18n setup