Even if it's P2P, you still need TURN servers because participants might be behind NAT that are hard to pierce through. This basically means you're proxying the full video traffic between both participants.
Setting up STUN and TURN is a real pain, though, if you run into any issues.
Actually, I'd suggest reading the entire document when you have time. For now, though, at least have a read of that section (it's short and sweet).
[0]: http://www.catb.org/~esr/faqs/smart-questions.html#beprecise
You won't be the first (nor the last, though if I can have the chance to improve that at some point I would love to) person to have struggled with it.