One day I was editing some recordings I did of a blackbird in my garden. I noticed that the different phrases it sang had certain similar sequences. There was a clear beginning, middle and ending part in each phrase. This sparked an idea for an experiment; What if i would take some of those phrases and cut them up in different parts in such a way, that I could create a sound object in Unreal Editor that stitched those samples together in random order, thus creating a new phrase each time the sound cue was played.
I ended up having 4 different beginning samples, 18 short middle parts and 4 ending samples totaling to just 1.4 megabytes of data . I made sure that every part was tightly cut, so the restitching of samples wouldn’t have noticible gaps of silence.
I then imported the sound files and created a soundcue;
Here you can see, every time the sound cue gets played, it starts with playing a beginning part, directly followed by up to 5 randomly picked middle part samples, and finally playing an ending part. For some of the middle parts I also added a very short silent sample, resulting in having a chance of that middle part being skipped in the phrase. This way a phrase could be shorter or longer, giving even more variety.
This is an example of how a little bit of work can result in a more believable, less repetitive audio experience, saving memory in the process. As opposed to originally only having 4 different phrases, you now have 19,200 ways a phrase can be built up.