You need to hold your head still, and then the software expands the bounding box to the full size of the screen. Everything I've described so far happens very quickly, and then slowly some of the words on the screen are translated and replaced in place. The lighting in my room is not very bright, and so far I've been unable to get full sentences to translate (just words, and just for a moment until I move my head a little), but I suspect it'd work better in high contrast situations (out in the sun, block text, etc.)
Tapping the side of Glass lets you switch between to/from English/Spanish.
Edit: Last thing, IMO, the largest downside to the UX right now is that when you move your face (even a tiny bit) the in-place translation disappears and you need to hold your face steady again to get the translation to appear. This would obviously make it pretty difficult to use in a real world situation (like, if you're walking outside – you'd need to stop and hold very still to translate a sign) but it's still super cool.