From Unify Community Wiki
Revision as of 19:19, 23 April 2009 by Jessy (Talk | contribs)

Jump to: navigation, search

Author: Jessy



This is a script that replaces most properties of Audio Sources. I feel that its controls behave more intuitively then those of the Audio Source component itself, and it extends functionality as well. I intend for it to be easy to switch from scripting and tweaking Audio Sources directly, to utilizing this script.

I have also created a simple derivative of this script, to control the volume of the Audio Listener.


If your Game Object has only one Audio Source, then this script will automatically control it, when attached. Otherwise, you can programmatically assign what the script controls via the audioSource property.

The Volume, Min Volume, Max Volume, and Rolloff Factor parameters perform the tasks that the similarly-named parameters of Audio Sources were intended to perform. If scripting these, where you normally would type "audio" in your code, use "Audio" instead.

Rolloff Threshold defines a distance from the Audio Listener within which sound will not get quieter over distance. (Unity normally fixes this distance to 1 world unit.) This can be especially useful for dealing with "extreme" scale characters. (Use large values for giant monsters, small values for insects.) You may find that, by using this parameter, you can avoid having to worry about changing Rolloff Factor to anything other than 1.

If Ignore Distance is unchecked, it allows stereo clips to fade in volume over distance (but they do not get positioned in 3D "sound space"). For mono clips, checking this is a more efficient method for achieving the same result as Rolloff Factor = 0.

Performance Mode makes the code execute faster, but the effects of all properties, other than Rolloff Threshold and Ignore Distance, are not automatically updated every frame. You can call the RefreshLoudness(), RefreshMinLoudness(), RefreshMaxLoudness(), RefreshVolume(), and RefreshRolloffExponent() functions, to make changes only when necessary, from external code. It makes sense to leave Performance Mode off when setting up the scene, and then turn it on after tweaking, if you don't need these properties to be extremely dynamic.

JavaScript - Audio.js

<javascript>// If there is only one Audio Source attached to this game object, // and there are no other instances of this script attached, // automatically link this script to the Audio Source. private var audioSource : AudioSource; function Awake () { if (GetComponents(AudioSource).Length < 2 && GetComponents(typeof this).Length < 2) audioSource = audio; } // automatically links the first attached instance of this script // to an Audio Source, if none were present @script RequireComponent(AudioSource)

var volume : float = 1; var minVolume : float = 0; var maxVolume : float = 1; var rolloffFactor : float = 1; var rolloffThreshold : float = 1; var ignoreDistance = false;

// determines whether the "Refresh" functions will run in Update(), // or must be called manually var performanceMode = false;

private var loudnessExponent = Mathf.Log(Mathf.Sqrt(10), 2); private var loudness : float; private var minLoudness; private var maxLoudness; private var rolloffExponent;

function RefreshLoudness () {loudness = Mathf.Pow(volume, loudnessExponent);} function RefreshMinLoudness () {minLoudness = Mathf.Pow(minVolume, loudnessExponent);} function RefreshMaxLoudness () {maxLoudness = Mathf.Pow(maxVolume, loudnessExponent);} function RefreshVolume () {audioSource.volume = Mathf.Clamp(loudness, minLoudness, maxLoudness);} function RefreshRolloffExponent () {rolloffExponent = Mathf.Log(.5, (1 / rolloffFactor + 1));}

private var audioListener : Transform;

function Start() { // These properties are not useful; // initializing them this way keeps this script as tidy as possible. audioSource.minVolume = 0; audioSource.maxVolume = 1; audioSource.rolloffFactor = 0;

// Unity only allows values between 0 and 1 for volume. minVolume = Mathf.Clamp01(minVolume); maxVolume = Mathf.Clamp01(maxVolume);

// ensure maxVolume is not less than minVolume maxVolume = Mathf.Max(minVolume, maxVolume);

// Negative values don't make sense for these. rolloffFactor = Mathf.Max(rolloffFactor, 0); rolloffThreshold = Mathf.Max(rolloffThreshold, 0);

// Having this in Start() implies that you won't be moving the Audio Listener // to another object during gameplay. var listenerComponent : Component = FindObjectOfType(AudioListener); if (listenerComponent) audioListener = listenerComponent.gameObject.transform;

RefreshLoudness(); RefreshMinLoudness(); RefreshMaxLoudness(); RefreshVolume(); RefreshRolloffExponent(); }

function Update () { if (audioSource && audioListener) { if (performanceMode == false) { RefreshLoudness(); RefreshMinLoudness(); RefreshMaxLoudness(); if (ignoreDistance == true) RefreshVolume(); else RefreshRolloffExponent(); } if (ignoreDistance == false) { var distanceToListener = Vector3.Distance(audioListener.position, transform.position); var rolloffBase = distanceToListener / rolloffThreshold; var positionalVolume = Mathf.Pow(rolloffBase, rolloffExponent) * loudness; audioSource.volume = Mathf.Clamp(positionalVolume, minLoudness, maxLoudness); } } }</javascript>

Why I find it necessary to use this

Tweaking the controls for Audio Sources in Unity is not as predictable as it could be. Here are the rogue properties of Audio Sources, with their descriptions as provided by the Unity Reference Manual, and descriptions as they have actually behaved in my tests.

(To be clear, volume is a multiplier for audio samples, which directly correlate with how far your speaker drivers (tweeters, woofers, etc.) are displaced from their resting position during playback. volume = 0 means that your speakers won't be displaced at all. Simplifying a bit, volume = 1 means that the audio will play back as loud as it was recorded, and fractional values will play the audio back as if it were recorded at a lower level. As samples get played back over time, waveforms are recreated, so saying that volume is a multiplier for amplitude is accurate enough that I will be using the term "amplitude" to mean 'distance from zero', later on this page.)

Play On Awake: If enabled, the sound will start playing the moment the scene launches. If disabled, you need to start it using the Play() command from scripting.

It actually means "Play On Enable". Yes, the sound will start playing the moment the scene launches, but it will also start playing if the Audio Source is disabled, and then enabled again, any time after the scene launches. I'm documenting this here, for the sake of completeness, but I am not providing a fix for this, because I don't think this is a problem, other than being incorrectly named and documented.

Volume: How loud the sound is at a distance of 1 world unit (1 meter) from the Audio Listener.

This is only true in the Audio Listener's YZ plane. A sound traveling along the Audio Listener's X is 3 dB louder than it should be, according to this description, and it only plays through one speaker, at least in my two-speaker environment. Obviously, this is due to the 3D positional audio algorithm, but the description of the property is not precise, and the algorithm is highly unrealistic along the Audio Listener's X axis. (Fortunately, it gets better if more axes are involved.)

To avoid distortion when within 1 world unit of the Audio Listener, along the X axis, it seems that all mono clips imported into Unity are reduced in level by 3 dB. The Unity 2.5 Audio Importer's "Audio Channels" box also assures that no clips that are converted from stereo to mono have any peaks above -3 dB. Unfortunately, although this is apparently not the case for everyone, on my system, Unity actually clips all samples above -3 dB to exactly -3 dB, which causes distortion for clips with loud peaks. (An appropriate solution would be to multiply all samples by 10^(-3/20), if the 3 dB boost along the Audio Listener's X axis can not be avoided.)

Also, while AudioSource.volume can accept values between 0 and infinity, values above 1 do the same thing as 1. That is to say, an Audio Clip will never play back louder than it was recorded. Boosting the level of audio clips can only be done outside of Unity, at least for now.

A further problem with the volume control is that it uses a "linear taper", which does not correspond with human hearing. For example, if volume is 0.5, instead of sounding half as loud as 1, it sounds about 2/3 as loud. If volume is 0.1, instead of sounding 1/10 as loud as 1, it sounds about 1/4 as loud.

Min Volume: The minimum value of the sound. No matter how far away you get, the sound will not get softer than this value.

Max Volume: How loud the sound gets at the loudest. No matter how close you get, the sound will never get louder than this value.

These variables actually have nothing to do with distance. Instead, they are just clamps for AudioSource.volume. The following odd behavior ensues, if Min Volume is greater than Max Volume:

Unity iPhone:

If Volume is less than Min Volume, then Volume behaves as if it were set to Min Volume. If Volume is greater than or equal to Min Volume, then Volume behaves as if it were set to Max Volume.

Unity 2.5:

If Volume is less than Min Volume, and Volume is less than 1, then Volume behaves as if it were set to Min Volume. If Volume is greater than or equal to Min Volume, or greater than or equal to 1, then Volume behaves as if it were set to Max Volume.

Like Volume, although Min Volume and Max Volume can accept values between 0 and infinity, values above 1 do the same thing as 1.

Rolloff Factor: How fast the sound fades. The higher the value, the closer the Listener has to be before hearing the sound.

"How fast the sound fades" is a good description, but the "before hearing the sound" part is inaccurate. A better way to put what Unity is going for is: "The sound will be as loud as if were 'Rolloff Factor' times as far away as it actually is, from the Audio Listener". However, that is not really what happens. The algorithm gets a little wacky, in an effort to balance this idea with the intention of a sound being as loud as AudioSource.volume at a distance of 1 world unit.

Sound has a friendly property: its measured amplitude decreases directly with distance from its source. (Take a measurement of amplitude, move n times farther away from the sound source than you already are, and a new measurement will yield 1/n times the original amplitude. This isn't true indoors, due to reflection and absorption of sound waves, but this ideal case is pretty accurate in open fields. Getting realistic indoor sound is not really within the capabilities of Unity's current audio engine.) So, using the realistic (outdoor) model, we could use the equation amplitude = volume/(rolloffFactor * distance), and accurately have the sound play back as if it were 'Rolloff Factor' times as far away as it is. The problem with this model, is that instead of the sound's amplitude being volume * loudness of clip at a distance of 1 world unit, it is volume * loudness of clip at a distance of 1/rolloffFactor world units. This codependence of properties is not particularly desirable, but unavoidable if adhering to this model.

In an effort to combine these two competing ideals, Unity uses the distance model of amplitude = volume / (1.0 + rolloffFactor * (distance – 1.0)). For mono audio clips, if rolloffFactor = 1, then the equation is physically accurate, and if rolloffFactor = 0, then there is some degree of positioning, despite the overall sound level not varying if distance from the Listener is altered. (See the description of what happens along the Listener's X axis, under 'Volume', above.) However, no other rolloffFactor values make sense, outside of a mathematical context.

For example, if rolloffFactor = 2, at a distance of two world units, you would expect the sound to be as loud as it would be at 4 world units, if rolloffFactor were 1. However, instead of the amplitude being multiplied by 1/4, as desired, the multiplicative factor would be 1/3. Tweaking rolloffFactor will probably be less intuitive, to anyone, than tweaking a good volume, minVolume, or maxVolume control, so I don't think it's as important to "fix" this property, as it is to fix the others, but I do believe that it can be improved.

Details of my audio control improvement methods

Volume: Although Unity's volume control is a precise multiplier of audio amplitude (from 0 to 1, anyway), the amplitude of a waveform does not correlate directly with human perception of loudness. Loudness perception is complex, but "experimentally it was found that a 10 dB increase in sound level corresponds approximately to a perceived doubling of loudness.". Therefore, it makes sense for a volume control to yield an amplitude that is...

exactly what is input, for a value of 1,

10 decibels below the input, for a value of .5,

20 decibels below the input, for a value of .25,


Amplitudinal difference, as a multiplicative factor, based on a difference in decibels, can be found with the expression 10^(dB / 20). Inputting the value of 10 dB results in 10^(10/20), which Mathf.Sqrt(10) in the code represents.

Raising this value to a base 2 logarithm gives exactly the behavior we want for a volume control. This expression would be (10^(1/2))^log[2]volume. However, the algorithm will execute faster in the form that I gave in the code, which is volume^log[2](10^(1/2)) (the RefreshLoudness function), due to only having to calculate a logarithm once, at runtime. Thanks to David Huntrods (dawvee on the Unity forums), for proof that x^log(y) = y^log(x):

y = x^log[x]y.

Therefore, x^log(y) = x^log(x^log[x]y).

Due to the property log[b](x^n) = n * log[b]x, x^log(y) = x^(log[x]y * log(x)).

That can be rewritten as x^log(y) = (x^log[x]y)^log(x).

Again, because y = x^log[x]y, x^log(y) = y^log(x).

Min Volume and Max Volume: These properties incorporate the same "human-corrected" calculation that Volume does. In addition, they function as described in the manual. As with Audio Sources, I do use them as clamps for Volume, but only if Ignore Distance is checked.

  • Although Unity restricts Audio Sources' Volume, Min Volume, and Max Volume properties, and my replacements for them, to a functional range of 0-1, my "skewing" would continue to function properly, above values of 1, if the restriction is ever dropped. (I use Mathf.Clamp01 in my code, right now, to make clear what is actually happening. I have always found it confusing for Unity to allow values greater than 1, as they don't reflect reality.)

RolloffFactor and Rolloff Threshold: Although I actually like what Unity is going for, in terms of its model for RolloffFactor, I definitely want to be able to set the volume of a sound at a certain distance, independent of the RolloffFactor setting, so a different model has to be used. Mine can be described like this:

Ideally, a sound becomes quieter by a fixed amount (about 6 decibels), as its distance from a listener is doubled. The sound will still become 6 dB quieter at a further distance, but that distance will be 'Rolloff Factor' times closer to any point outside of Rolloff Threshold than it would be in ideal reality.

Keep in mind that Rolloff Threshold can not be zero, as sound measurements must be taken at a certain distance from a source. This distance can be infinitesimally small, but can not be zero. Given that AudioSource.volume is a linear amplitude multiplier, my equation is as follows:

audio.volume = loudness * (distance/ Rolloff Threshold) ^ log[1+ (1 / Rolloff Factor) ](1/2), where loudness is the result of the volume equation detailed above.

Let's break this down. Assuming Volume = 1, Rolloff Threshold = 1 (which imitates the behavior of the Audio Source's rolloff model), and Rolloff Factor = 1, then we can simplify the equation to

audio.volume = (1/2) ^ log[2]distance.

This can be read as...

As distance from the Audio Listener is doubled, outside of the range of 1 world unit from the Listener, the amplitude of the audio signal will decrease by half.

(Halving the amplitude results in a decrease of 20 log(1/2), which is about - 6 dB.) At a distance of 1, the amplitude is 1. At 2, the amplitude is 1/2. If we instead wanted the amplitude to be 1/2, twice as close to 1 as 2 is (1 + (2-1)/2, a.k.a. 1.5), then we would substitute 1.5 into the logarithm:

audio.volume = (1/2) ^ log[1.5]distance

The general case for this is audio.volume = (1/2) ^ log[1+ (1 / Rolloff Factor) ]distance. However, when distance = 1, then audio.volume will be 1, no matter the Rolloff Factor. To change it so that audio.volume = 1 at a distance of Rolloff Threshold, then we first divide distance by Rolloff Threshold, before computing the rest of the equation. This yields...

audio.volume = (1/2) ^ log[1+ (1 / Rolloff Factor)](distance/ Rolloff Threshold),

which, when the Volume property is accounted for, can be modified to the final form seen above, using the property of logs + exponents that was detailed under Volume.

To elaborate on what I mentioned in the Instructions, Rolloff Threshold is probably best used to match the overall scale of your game. Rolloff Factor can be used to emulate the sonic effects of different environments, in an extremely simplified fashion. Highly reflective rooms, like empty basements with concrete walls, would be more realistic with lower Rolloff Factor values. Higher values would be appropriate in environments with more sound absorption. An attic filled with fiberglass insulation is an excellent example of such a space. When Rolloff Factor is set to 1, "realistic outdoor behavior" will be modeled. (This is also true of the Audio Source's Rolloff Factor.)

  • Be aware that mono clips will exhibit the "+ 3 dB along the X axis" behavior, despite the use of this script. There is no way to code other types of positional audio, in Unity, because Unity does not provide access to individual audio channels.

Ignore Distance: Although, when unchecked, this does allow the volume of stereo clips to be automatically faded over distance, because the script does not have access to Unity's internal audio smoothing algorithms, you run the risk of introducing "crackling" artifacts into your audio, if you change the position too rapidly. This is due to samples jumping rapidly to other values, drastically changing the slope of the waveform momentarily.

Personal tools