A Virtual Voice is a Voice that does not use any sound generation resource for playback: only the update of parameters and time are managed.
When playing a Cue with a player that has the "CRIATOMEXPLAYER_RETRY_VOICE_ALLOCATION" option set on the program side, if a Voice cannot immediately be secured when starting the playback, or if the Voice was stolen because it had a low priority, the playback information is kept as Virtual Voice.
The Virtual Voice tries to secure the sound resource during the server process and changes to a normal Voice if it succeeds. The actual sound processing is then performed.

On the tool side, you can also configure a voice to become virtual when its volume becomes "0", i.e. if it becomes silent.
Indeed, it would be wasteful to use sound generation resources to simply play silence (e.g. when the volume becomes "0" due to distance attenuation when using 3D audio positionning.)
In this situation, the virtualization of the voice saves resources.
The configuration of the Voice virtualization is done in Voice Behavior of the Cue in the tool.