The audio offset controls where you are in the series of people. The larger your offset, the farther along you are in the chain, the more people will be ahead for you to hear, and the fewer people will be behind to hear you. If you're with a group of people, figure out now who will be at which position. A good place to start is people at 0, 10, 20, 30, 40, etc. Numbers over 100 generally won't work because the server buffer isn't large enough. Eventually, we're planning to tune this so delays can be much smaller.