Computer Games Sound Effects
Recording, Postproduction and Existing Database
Bartosz Zi´ołko, Martyna Gromotka and Mariusz Zi´ołko
Department of Electronics, AGH University of Science and Technology, al. Mickiewicza 30, Krak´ow, Poland
Keywords:
Video Games, Signal Processors, Audio Edition.
Abstract:
The paper describes the process of building a new database of sound effects recordings for computer games
and the first version of such product. Ways of applying signal processors for postproduction is described, as
well as differences in audio edition for lms and games. Some aspects of using sounds in games are also
mentioned as well as the rst version of the list of possible tags of the audio files in the database. Both the
language of the tags and the datatabse will be substanially enlarged.
1 INTRODUCTION
An important factor at work in the audio creation
for games is dynamics of the signal (Marks, 2001;
Collins, 2008). It means the difference between the
worst and best possible level of the acoustic signal to
output or input of the device (Sztekmiler, 2003).
In real cases it is very difficult to record sounds
with a high dynamic range. Playback and transmis-
sion of acoustic signals have their limitations as well.
Some very soft sounds will be inaudible at the play-
back, and those with a higher level, will be too loud
or distorted. Each recording or a reproducing device
has a certain level of noise and maximum output level
beyond which sound becomes deformed. Because of
the dynamics limitations, the acoustic signal devices
should be maintained above the noise level and below
the threshold value of voltage distortion.
Another important issue in case of games is that in
contradiction to audio edition for films, sound effects
have to be ready for non-linearity and mixing during
the course of the game. The effects can be stopped in
a middle or overlaid on each other if a player makes
some actions. In case of overlaying, a simple correla-
tion can be not appropriate as typically sound related
to events in a game but can be be more important then
background music. Some sound effects have to be
played in a loop, but in a way that the end point is not
hearable.
2 SOUND EFFECTS
PROCESSING
The dynamics processors were helpful in building the
multimedia database. Depending on the level of the
input acoustic signal they influence or reinforce the
acoustic attenuation of the track in which they are
placed. Dynamic processors are used, not only be-
cause of the possibility of signal distortions, but also
to provide a listening comfort. If the dynamics of the
recorded material is reduced, manual changes of the
volume playback is not needed. In the opposite case,
the quieter parts of the volume could be not noted at
all, because the sound would be inaudible.
On the other hand, it could appear that the par-
ties, in which an acoustic signal reaches a high level,
are too loud. With these devices, one can manipulate
the value of the effective signal, which largely deter-
mines the subjective loudness by the recipient. For
the higher root mean square (RMS) value of the sig-
nal, the audio seems louder, even though the peak re-
mains the same. We used mainly following dynamics
processors: a compressor, a limiter, an expander and
a gate (Lyons, 2004) while building the database.
2.1 Compressor
The compressor reduces the signal dynamics. Typi-
cal target of its usage is to raise the RMS of a signal.
However, first, the compressor attenuates the signal,
which exceeds a predetermined level. In this way, the
entire recording can be strengthened without exceed-
ing 0 dBFS peak (in the case of digital recording),
223
Ziółko B., Gromotka M. and Ziółko M..
Computer Games Sound Effects - Recording, Postproduction and Existing Database.
DOI: 10.5220/0003977702230226
In Proceedings of the International Conference on Signal Processing and Multimedia Applications and Wireless Information Networks and Systems
(SIGMAP-2012), pages 223-226
ISBN: 978-989-8565-25-9
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
which would override the sound track. Therefore, the
compressor reduces the difference between the loud-
est and quietest parts of the path. The signals are am-
plified in a proportion corresponding to the compres-
sion ratio. Parameters that can be adjusted in most
compressors are (Sztekmiler, 2003):
Threshold is responsible for the moment at which
the processor is on or off. Threshold sets the level
of the input voltage, above which there is signal
loss. The signal below is unchanged.
The size of gain reduction is determined by the
compression ratio
k =
U
in
U
out
(1)
where k is a coefficient of compression and U is
the size of input/output voltage above the thresh-
old in dB. This parameter indicates how many
times the input voltage level is greater than the
voltage at the output, e.g. 2:1, 10:1. This means
that a proper amplification beyond the level will
be reduced by two or ten.
The rate of response (attack) describes how
quickly after the crossing of the size of the in-
put threshold voltage level compressor will start
its function.
Releasetime to the previous settings of earlier am-
plification, after going back beyond crossing the
level set by the parameter threshold.
Make-up gain enhances the output signal. It is
used because of the reduction strengthening of the
compressor.
Knee is responsible for the part of the transient
characteristics of the processor in the activation
threshold. Soft knee is used to make compression
more natural and less audible, and sharp one for
strong compression.
Peak-RMS - compressor reacts to the RMS or
peak.
Short attacks of the compressor is usually used for
transient signals, such as explosions. In case of long
ones, the compressor would not manageto pass a fast-
changing signal, so its sound would change dramati-
cally. It could be heared as a significant beginning of
the transient signal attenuation (the impact moment)
and will only sustain itself. However, sometimes it
can be done consciously to have this effect. Param-
eter Peak / RMS should be set also depending on the
type of a signal. Selecting peak is appropriatefor tran-
sient signals. Then the processor will respond to the
peak value, and not an effective signal. If the sound
is continuous and there are no significant changes in
the level, RMS should be chosen. The processor then
takes into account the effective value, and its effect
will be less noticeable and more natural, which is of-
ten desirable for this type of signals.
Increasing the level of silent parts will increase the
noise level recorded track, so the compressor, as well
as other dynamics processors, should be used con-
sciously, knowing precisely the effect one wants to
achieve. All processes of denoising audio material
should be performed before using the compressor, be-
cause it will emphasise the unwanted sounds in the
recording. Dynamic processors are helpful tools for
editing audio, but it is not reasonable to use them too
much because they cause flattening of the sound and
artificial sound recordings.
2.2 Limiter
The limiter is a dynamics processor, having similar
characteristics to the compressor. Its characteristic
feature is that the compression ratio tends to infinity,
and the attack is as close to zero as possible. They are
the hardcoded parameters. Limiter is used primarily
to prevent the signal from clipping, reacting mainly
to the peak values of the signal. Thus, it does not al-
low to go beyond the level of 0 dBFS. This means that
after applying the limiter, output signal level will not
increase regardless of the level of input signal. The
processor allows to adjust the following parameters:
Threshold - after crossing the threshold, limiter
starts to operate at time close to zero.
Release time to the original gain.
The limiter should mainly protect from distortion
of audio, activating relatively rarely. Frequent cutting
of the signal by a limiter would have a significant im-
pact on quality, because the nonlinear deformations
will appear. What is more, the dynamics will be sig-
nificantly reduced and the sound recordingcan be per-
cepted as unnatural.
2.3 Expander
An expander works oppositely then the compressor.
It reduces the amplification, but of the quietest parts,
with a level below the
threshold
. The same param-
eters can be adjusted, as with the compressor the
attack is responsible for the delay of the expander,
and the release for the late inclusion of gain reduc-
tion when signal levels drop below the threshold. Too
high value of the response time can result in attenua-
tion of high, rapidly rising level, which will be heard
as “ragged”. The expander is used primarily to mute
SIGMAP2012-InternationalConferenceonSignalProcessingandMultimediaApplications
224
noise or unwanted sounds that occur between the use-
ful material.
Noise gate is a version of expander with ratio pa-
rameter set to 1: . Then, no input signal, which falls
below the level set by parameter threshold, cannot be
passed to the output. Thus, the phonic circuit closes.
The main goal is to get rid of unwanted sounds, such
as glitches, breaks or noises which are heard in the
intervals between the wanted sound. In addition to
the gate parameters such as threshold, response time
(time from the moment of crossing the threshold to
gate opening), the recovery time (time from signal
drop below the threshold to closing the gate). One
can also adjust the hold time, which is responsible for
the opening of the signal processor after falling below
a preset threshold.
This processor is better than the standard denoiser
built-in the most of audio editing programs, because
it does not degrade the signal, and “cut” only the un-
necessary parts. If noise interferes with audio during
the useful sound, the gate is useless.
2.4 Summarisation
All of these dynamics processors not only attenuate,
suppress or do not pass the signal, but also signifi-
cantly affect its sound. With their help, entirely new
sounds can be created, sounding very different from
those played back immediately after recording. This
is not always a desirable effect, but usually the cor-
rectly configured processor improves and enhances
the sounds from the recording material.
Dynamic processors can operate not only across
the whole band, but also in certain frequency sub-
bands. Band compressors divide the acoustic spec-
trum into areas in which all the parameters can be set
independently. Thus, each band is subjected to a dif-
ferent compression allowing more precise processing
of the dynamics.
Equalizer can be also useful to adjust the fre-
quency of acoustic signals. It can change the volume
for a specific frequency. It consists of band filters,
so that it is possible to amplify or to suppres of cer-
tain frequencies. Therefore, the timbre is changed.
Equalizer is used, among others in shaping the sound
recordings and correction of problematic frequencies.
Editing of the sound effectswas performed using a
digital audio workstation (DAW) developed by Stein-
berg - Cubase (Childs, 2007). It has a wide range of
tools for recording and editing, mixing and mastering.
Depending on the type of the edited signal, dif-
ferent signal processors were used and different fre-
quency corrections. It was possible to set the dynam-
ics in order to get rid of much of the reverb, which
is especially disturbing in impulse sounds. More
difficult to edit were the continuous sounds, due to
less possibility of interference in the recorded sam-
ple. Both graphic equalizers and parametric ones
were used to obtain the desired effect.
Short duration paths were modified to allow to
loop the sounds for the continuous playback of one
track without noticeable discontinuity between the
end and the beginning. It was necessary to extract
additional fragments, where the sound was quite con-
stant. If there would have been too many changes or
pulses that stand out, the soundtrack would be some-
what rhythmic and predictable. Then the loop would
have a noticeable effect, which we wanted to avoid.
In the last stage normalisation has been made by
finding the samples with the highest peak and setting
the level of the track, so that the highest value reached
almost 0 dBFS. Thesignal should notcross it, because
then, there would be a digital clipping. This is why
a limiter was plugged at the output of the circuit to
protect from the distortion.
There is no clear existing standard, which would
inform what level of the effective value of the output
path should have a sound effect for a computer game.
The RMS value depends mainly on the scale of us-
ing dynamics procesors during mastering, in particu-
lar – the compressor. The smaller the difference in the
level of quiet and noisy signals, the greater becomes
the effective value of the signal. For music, these val-
ues permanently grow (Owsinski, 2006). After 2000,
they reached -10 dBFS, and for some songs even -5
dBFS, especially in popular music. For classical mu-
sic, where the dynamics of the song is a very impor-
tant element, the effective value is approximately -30
to -20 dBFS. In this case the dynamics is not as flat
as in typical entertainment music. We decided that
the RMS for our sound effects will oscillate in similar
limits as is the case of classical music.
3 COLLECTION OF SOUND
EFFECTS FOR GAMES
The samples were stored in a collection of 305 sound
effects, which are 45 minutes from paths which to-
gether lasted 3 h 10 min. Overall, post-production
of sound effects for the collection lasted about 70
hours not including the process of recording samples,
which lasted several days. One of the major obsta-
cles was tiring of hearing, therefore, quite frequent
breaks were necessary. After a two-hour work of lis-
tening to the sound effects, the perception is worse,
so one can not properly perform the tasks assumed.
The collection is just the preeliminary one to establish
ComputerGamesSoundEffects-Recording,PostproductionandExistingDatabase
225
proper work routines, methods and to allow to work
on soundtracing algorithms for games. The collection
will be substantially increased in the following two
years to reach over 10 hours of sound effects.
The file format is an uncompressed PCM with a
sampling rate of 44.1 kHz and a resolution of 16 bits.
Files names of the particular sound effects give an ad-
ditional information about the sound: I - pulse sound,
C - continuous sound, L1 - possibility of looping, L2
- possibility of looping without impression of repeti-
tive sound, K - effect allowed for kids, A - effect for
adults only, F - effect good for funny games, P - effect
for positive games (Zynga type), H - effect for horrors
and mysterious games.
4 CONCLUSIONS
The process of preparing sound effects for games was
described. An existing and growing collection of such
audio files have been presented as well. The collec-
tion is a result of just started project. The most im-
portant aspect of preparing sound database for games
besides of quality is their usefulness in non-linearpro-
cessing, because sound effects have to be mixed a lot
and played in very changeable orderdepending on the
decisions of a player.
ACKNOWLEDGEMENTS
This work was supported by NCBiR grantfor RAYAV
project.
REFERENCES
Childs, G. W. (2007). Creating Music and Sound for
Games. Boston.
Collins, K. (2008). Game Sound. An Introduction to the
History, Theory, and Practice of Video Game Music
and Sound Design. London.
Lyons, R. G. (2004). Understanding Digital Signal Pro-
cessing. Prentice Hall PTR, Upper Saddle River, NJ.
Marks, A. (2001). The Complete Guide To Game Audio. For
Composers, Musicians, Sound Designers, and Game
Developers. Lawrence.
Owsinski, B. (2006). The Mixing Engineers Handbook.
Boston.
Sztekmiler, K. (2003). Podstawy nagło´snienia i realiza-
cji nagra´n. Podre¸cznik dla akustyk´ow (in Polish) Eng:
Fundamentals and implementation of sound record-
ings. Handbook for acoustic engineers. Wyd. Naro-
dowe Centrum Kultury, Warszawa.
Figure 1: Recordings of three sound effects: birds (in the
end intentionally fade out), a camera and someone walking
on leaves and their mixture for an event in game of a person
walking on leaves, taking photos of birds and going away
afterwards.
SIGMAP2012-InternationalConferenceonSignalProcessingandMultimediaApplications
226