Yeah, nice idea.
But then: How to recognize/designate a/any language properly?
Full name as in
https://en.wikipedia.org/wiki/List_of_o ... _territory
2-letter country code as you suggested or as in TLD
https://en.wikipedia.org/wiki/List_of_I ... uage_codes
or 3-Letter
https://en.wikipedia.org/wiki/List_of_I ... uage_codes
Or following subtitleedit into NLLB:
https://github.com/thammegowda/nllb-serve
Also can be a pain when there is serbian or kazakh in either latin or cyrillic, not to mention other countries and their multiple scripts...
Well, the commonly used 2 letters would be probably very prone to errors. Still, nice idea. Perhaps a subset (of safe country codes) would be feasible?