In an embodiment, a server obtains a user-specified mixing preference (USMP) for an access terminal (AT) in a communication group that is different than a default mixing rule for the group communication session. The server grants non-exclusive floors to a subset of ATs in the communication group, receives media frames from each AT in the subset and generates an output frame by selectively mixing the received media frames based on the USMP, after which the output frame is sent to the AT. In another embodiment, the AT obtains the USMP and receives unmixed versions of the media frames and then selectively mixes the received media frames at the AT based on the USMP. In another embodiment, an AT detects speech and requests the floor, and the AT buffers speech while waiting for the floor request to be granted.