New School Blues Dev. Diary #20: Audio Fidelity

For those who might not now and are wondering why some games run slower then others, often it’s because the game needs time to generate all the assets needed on screen.  Think of it like delivering food for a party.  Delivering one pizza pie would take you one trip, but delivering 50 pizzas or one giant pizza would mean multiple trips in and out of the kitchen, and that takes more time.

One way to avoid this is to strive to have assets be as small as possible without losing video or audio fidelity.  Please understand this doesn’t mean things are unpolished, it just means for the sake of getting the game to run smoothly, we trim what we can to keep it both looking nice and playing fine.  It’s an age old balancing act that every developer, no matter how big or small, has to deal with.

Audio assets work the same way in that always going for the highest quality can really tax your game’s performance.  Fortunately, Ryan has dealt with this issue before and already had some samples of different audio qualities for us to choose from.  Since our script was simplified and only one character had a speaking role, we were able to choose a higher level of fidelity without slowing down the game.  This works out great because high quality sound for the narration is key to our game maintaining that storybook feel.

Take a listen to the following tracks to see just how much compressing the size and quality of a vocal track affects how it sounds.  While different sizes, they are all the same length and line reading.

- The full quality one (44100sr), file size: 143k

- The medium quality one (16000sr)  file size: 12k

- The low quality one (8000sr), file size: 8k

Obviously Ryan is able to tweak things around to some extent and give us fidelity levels in between these examples, but they serve as nice starting points for devs to refer to.

Tomorrow we’ll talk a bit more on sound file types, and why having the correct one for certain audio assets is more important then you might think.