Hi, I would like some advice on sound quality for a 3d fps-like game. My engine uses OpenAL and exposes features like doppler effect, EAX, sound volumes/modifiers, occlusion, selecting random sounds from a group (by distance, in sequence, randomly etc). Sounds are stored in WAV or OGG, OGGs can be streamed. I don't have anything like a caching system for sounds, and I don't want to add any, keep things simple. What bitrate/frequency I should use for what effects? For example, should I use a higher bitrate for louder sounds (like gunshots)? If there's little audible difference, is it ok to assume it sounds well on all types of hardware setups? Also, what type of quality is generally acceptable these days? What quality would you expect from a game? I'm not really a sound guy, my question is kinda vague but I'm interested in general advice and some tips, anything really. Things I should watch out for? Thanks. :)

44100 16 bit resolution is pretty standard for .wav format quality. I definitely wouldn't go below that with your raw samples, simply because manipulating the file is best done in that format prior to putting it into a lossy format like .mp3 or .ogg. In the case of OggVorbis, or Mp3 files I wouldn't personally recommend a bit rate below 128 kbps, since you'll notice a lot of degradation starts below that (around 96 kbps its -real- noticable) but I prefer 160 kbps as a good median between quality and file size.

I think you're right on the money - different types of effects will require different khz rates, and careful discretion of mono vs. stereo. For example, any kind of sound that will be positioned in 3D (gunshots, creatures) should be in mono; other sounds like gigantic explosions or sounds for your first person character (eg. reloading) can be stereo for enhanced clarity/effect.

As far as quality goes, in my opinion it's the frequency range that should make the difference, not necessarily loudness. For example, a sound with absolutely no high frequencies (rumbles, ambiences) can be in 22khz, sometimes even 11khz really; not much point in preserving high frequencies that aren't even there. Otherwise, 44khz is a good call, especially when in doubt.

In my opinion, the games with the most successful sound design are the ones where there's several fields of sound operating simultaneously and independently - room ambience, object ambiences (a computer in the distance, an alarm; they don't even need to exist graphically for you to put them in), nature ambiences (wind, water, etc.), creature ambiences (low breathing/grumbling), 1st-person character sounds and foley (footsteps and clothes rustling). On top of that you'll have the expected gunshots, explosions, voices/creatures, UI and so on, but really it's those ambient sounds that make the scene come alive even when technically nothing is happening. It's what gives the player an emotional reaction to a new area he's just entered, and can have an incredibly powerful subconscious effect.


It might be a good idea to hack into the sounds directory of your favorite games and check out how other people have organized and prioritized their sounds, which ones are mono/stereo and which ones are 44/22/11. I'm pretty partial to the Half Life 2 sound design myself, some beautiful stuff there :-)


------------------
Wilbert Roget, II
RogetMusic.com

Unless you're writing for previous generation consoles, or handhelds, I'm not sure it's worth the effort trying to figure out what sounds will have 32 kHz vs 44 kHz sample rate. The only difference that makes a difference is mono vs stereo. Sounds at 22 kHz sample rate lack a certain sparkle, almost no matter what the sound (footsteps, gunshots, car engines -- doesn't matter). This should be obvious with a good pair of studio monitors, or a good pair of headphones. (Although the $7 headset that came with your MP3 player probably doesn't count as good :-)

When it comes to "sound caching" -- you have to pre-load certain sounds, because it'll take too long to load the sound from disk each time the user shoots a gun, say. The delay will be quite noticeable. I would call this a "sound caching" system.

AND

Some soundcards and drivers don't like mixing sample rates one after the other. Playing 22KkHz gunshots over 44.1kHz music with 22k ambience can create some odd artifacts (also depending on what sound engine is being used). Unless memory footprint is a problem I always push for uncompressed SFX. All positional audio is mono, certainly when dealing with multiplayer games. If it's truly not an "emitter "and there is only one "listener" (like menu and interface sounds) then go for stereo, but a gun reloading should also be considered positional in a networked game.

Definitely check out FEAR COMBAT (which is free) for a fantastic aural experience and great use of ambience and positional sound (sorry, no music!). Never before has audio CHANGED the way I played a game than this one. Bigger games go crazy, but the small skirmishes are tense and challenge you to use more than just a quick aim and good reflexes. You HAVE to listen to everything from debris on the ground to someone reloading in the next room.

Fun!