This is the model that will be used. This is optional and will default to the correct model for the voiceId.