Pretty good video in my opinion understanding how speech to speech model would work and possibly how openai realtime works if they are using a similar architecture