I don't think you can get away without using ANY text in your app. Specially in the set up of user preferences.
To make it easier to the user, you can take a 2 step approach.
Screen 1 - asks about User's native language. This will have to be a dropdown (Standard practice is Using English Text). Using Flags may or may not work depending on how many languages you wish to support. For example, India is one flag but over 20 languages. If you are only supporting national languages, Using flags might work.
Now that you have the language that the User speaks,
Screen 2 - will be in that language asking what language they wish to learn and it's the same dropdown. People will be familiar with this dropdown the second time and will be easier for them.
I understand that you are trying to do something novel here by not using text and such.. Would love to hear what your ideas are. Often best practices are ones that are standard and people are used to them. Eg. Check out Rosetta stone. They also use an English Dropdown.