Sound Designer Graham Gatheral recently posted a great tutorial on managing audio memory using the level streaming functions in UDK. In the post Graham steps through how to set up streaming volumes so that audio can be dynamically loaded and unloaded. This allow for memory to be conserved and not entirely front-loaded into RAM. The video below is a demo of the system showing a player entering different regions of a level and triggering audio load/unload events. All of this occurs while playing only important region specific audio. Fire up some Kismet and check out the tutorial here!
The benefit of placing audio assets and Kismet sequences in streamed audio-only levels is it gives you very tight rein on what sounds are loaded into memory at any one time – thereby helping you to stay within budget and not exceed your maximum number of channels. It also keeps audio separate from the rest of the game, which allows for greater control over your assets/implementation, and means that the files you need to work on are less likely to be locked. This tutorial will show you how to put your audio assets in designated audio levels, and control the loading and unloading of these assets using level streaming volumes
