"Every language" sounds like a massive stretch. Then again I come from a country of 200+ different languages, so maybe my expectations are a tad higher than most.
Unfortunately, given the reliance on machine translations, I don't expect it to be a useful learning tool for most of those languages any time soon.
As an aside I'm curious about your voices, especially for highly tonal languages like Yoruba that also don't necessarily have a lot of easily trainable material available.
While we offer people the ability to do machine translations, there's absolutely no reliance on them. We actively advise against using machine translations when onboarding new users
It's a good point about the "every language" part. That's really the goal of the project - we've been adding about 20 languages a month since we started, and don't plan to stop until we support every language. However we're not there yet, give us a few more months to live up to that ;)
We have some ideas for languages that don't have widely available TTS, but we're not ready to share anything quite yet.
Ah! To be clear I was going off of other comments, and also omitted some of the steps in my logic. Basically I mean that the project seems to rely on third-party-sourced translations of material which are themselves often machine-translated, and that support for new languages seems to be added through machine translation (though I suppose these days we call it AI instead). To be clear this is as opposed to something like Duolingo, which hires humans to translate its available material.
Yes! I mean, we won't supply the content for all 7,000 languages, but we can make the core learning process and core pipeline language agnostic, so you could learn any language.
With the exception of sign languages of course, as those will require their own processes and pipelines.
Unfortunately, given the reliance on machine translations, I don't expect it to be a useful learning tool for most of those languages any time soon.
As an aside I'm curious about your voices, especially for highly tonal languages like Yoruba that also don't necessarily have a lot of easily trainable material available.