Snd Wiki Changes
Posted 13 July 2006 - 10:05 AM
Added a Tutorials/Resources section, and copied in the contents of the Active: Sound->Active Sound Tutorials forum. I'm requesting that these threads be moved to the warehouse, and the forum deleted.
Kafros, you contributed these links - one of the links is dead (Simple intoduction by Silexz) - there's still a site there, but I couldn't find the tutorials on it. If you still think the tutorial is useful, and can find it, please add it to the Wiki.
This is the only page that's actually new to the Wiki. Every other department had a Docs page, and I was jealous The main thing here is Workflow - I got the original from the Active: Sound->Active Music Guidelines forum, but I've added to it.
Music Asset List:
Big change here - the only thing that's the same is the format requirement. I copied all information from the Assets spreadsheet into the appropriate Issues in the Bugtracker, and removed reference to it in the Wiki (actually, its crossed out at the moment, to be removed after review). I also moved in some guidelines ATeX had created for music, and links to X-Com MIDI files.
Sound Effects Asset List:
Least changes here - I just removed the italics on the Mastering Level guidelines that I added last week - since there were no objects, I guess this makes them Accepted.
I'm also requesting the Warehousing and removal of the Active Sound Guidelines forum - everything there is now on the Wiki. I'll send PMs to the appropriate people for this.
(I didn't touch the Archived Music List)
Posted 13 July 2006 - 03:37 PM
one thing I like to do when I make a sfx is to make sure there's no clipping.
Some of the values listed for max dB may have clipping. especially if your around -1dB !
I've been using soundforge and it has a tool called "normalize" I've been setting that to
around 90% (-10dB) to avoid any clipping. I think Audacity has a similar tool/feature.
I think this compresses the signal to the same level, which is a little different than
sound volume. but they go hand in hand with the signal level.
I've been trying to make all my samples the same level. around -8dB to -12dB.
Even soft ones. I figure the program can adjust for overall sound level in some control panel.
If the sample is too low, its hard to hear (especially compared to other sounds)
and if its too loud, it clips.
I think it would be better to have all samples at about the same level and have a control applet
adjust the sound SFXs as a group.
one other technique pointer, I also like to have a quick, very short duration, fade in and fade out, to avoid any spikes. Without the fade in / fade out there can be a very short spike that detracts and sounds horrible compared to the rest of the sample. The way I see SFXs implemented, I dont think they need to precisely loop, like in a song... All of the samples are not the same duration/length, either... so they can't really loop like a song.
Posted 13 July 2006 - 04:24 PM
Here's the best reference I was able to quickly find on the subject:
Here's a pretty good directory for this type of thing:
The normalize tool in Sound Forge (when normalizing to peak value) simply adjusts volume - it's not a compressor, which dynamically reduces the volume of input when it goes over a set level. Most importantly, its not possible to clip while using it - here's a clip from its Help file:
"From the Process menu, choose Normalize to raise the volume of a selection so that the highest sample level reaches a user-defined level. Use normalization to ensure you are using all of the dynamic range available to you without clipping. "
So, when normalizing to peak value, its pretty straightforward, and I specifically mentioned peak value the Wiki. It gets a bit more complex when normalizing to RMS, which does apparently use some compression techniques:
"When you normalize using average RMS power, Sound Forge will raise the average RMS value of the sound file to a value you specify. This is helpful for matching the apparent loudness of different recordings."
Here's a good explanation of compression (which is about as confusing as pointers and recursion, in my opinion):
Theoretically speaking, for the best resolution, we should be setting peak levels of everything near -0.1. However, since most initial sounds are recorded at lower levels, that higher resolution isn't really meaningful. If things are at too low a level, then in addition to the problem of simple quietness that you mentioned, you start really losing resolution. I proposed a lower cap of -18 dB based on statements from various Mastering engineers - I can't find the references right now, though.
So, its fairly arbitrary where the peak levels are between -18 and -.01.
Since we have that area to play in, I think its useful to have relative values within a logical set of sounds. The levels I set for the different Morlock samples help to demonstrate their intended use, which I think will be helpful when they're first incorporated into the actual game - they communicate what I was thinking in design. Similarly, it might be useful to set the smallest UFO's sound at -12, and have the largest near -1. Its a given that levels will be changed when everything gets incorporated, but this seems like a way to get a little head start.
I'm open to continued discussion, though.
Edited by LfO, 13 July 2006 - 07:05 PM.