I know it is an old topic, but I'll reply anyway.
I was able to pull this off by using a parallel script that controlled the conversation between four actors, not just two. I checked for the players x, y coordinates to make sure they were in range first (so the text bubbles could be displayed), then allowed the parallel script to show the conversation. The conversation appeared to the player to start in the middle, but I looped it so nothing was actually missed.