So while it's useful to have language specific character recognition, I believe this tool will best serve it's users when instead of using a default language it'll use the script/Latin training set as default. This has been trained on latin characters (plus extended latin characters) and should be able to read text on screenshots very accurately, regardless of language (assuming it's one in a latin script).
I also think it makes sense including other scripts as well. These are great language agnostic ways to get good accuracy. If required people can still switch to a language specific model.
Additionally teseract 5 has been released. So might be worth updating to that.
So while it's useful to have language specific character recognition, I believe this tool will best serve it's users when instead of using a default language it'll use the script/Latin training set as default. This has been trained on latin characters (plus extended latin characters) and should be able to read text on screenshots very accurately, regardless of language (assuming it's one in a latin script).
I also think it makes sense including other scripts as well. These are great language agnostic ways to get good accuracy. If required people can still switch to a language specific model.
Additionally teseract 5 has been released. So might be worth updating to that.