If the app has to take a photo at the user's behest everytime a word needs to be translated, it's going to be very clunky very fast.

The whole point of technologies like Glass is that they should be as unobtrusive as possible and just work by themselves when you need to.