Audio

Sound works the way it does in Unity: each thing that makes sound has an Audio Source, and one Audio Listener is the player’s ears. The engine is miniaudio, compiled into Tallow, so nothing needs installing; it finds PipeWire, PulseAudio or ALSA by itself when the first sound plays. A game that plays no sound never opens the sound card.

Sound files are WAV, OGG, MP3 or FLAC, kept anywhere in the project. Double-click one in the Project panel to hear it.

Audio Source

Add one with GameObject > Audio Source, or Add Component > Audio. Pick an Audio Clip and press Preview in the Inspector to hear it without entering play mode; outside play mode the Scene view’s camera is the listener, so you can fly around a source to judge it.

With the source selected, the Scene view shows its minimum and maximum distance as spheres.

let source = GetComponent<AudioSource>();
source.Play();                              // the Audio Clip
source.Play("Sounds/alarm.ogg");            // a different one
source.PlayOneShot("Sounds/step.wav");      // on top of whatever is playing; any number can overlap
source.Pause(); source.Resume(); source.Stop();
if (source.IsPlaying) { }

Length and Time give the clip’s length and the playback position in seconds; setting Time seeks.

Audio Listener

Put an Audio Listener on the camera or on the player’s head. If the scene has none, the first Camera is used, so a simple game needs nothing.

Sounds without an object

Audio.PlayOneShot("Sounds/explosion.wav", position);     // at a place in the world
Audio.PlayOneShot2D("Sounds/click.wav");                 // no position: interface sounds

These play once and are freed when they finish.

Headphone 3D (HRTF)

HRTF is off until you turn it on; until then direction is ordinary left-right panning. Turned on, each positioned sound is filtered through the responses measured at the two ears of a dummy head for the direction it comes from. That gives the brain what it uses to place a sound in real life: the sound reaches the nearer ear first (by up to about 0.7 milliseconds), is louder there, and is coloured differently depending on whether it is in front, behind, above or below. On headphones a sound behind you sounds behind you.

Audio.Hrtf = true;       // headphone 3D; can be switched while sounds play
Audio.Hrtf = false;      // ordinary left-right panning (the default)

It is made for headphones. Over speakers the two ears each hear both speakers, which undoes the effect and can make sounds seem coloured, so put the choice in your options menu (“Headphones” and “Speakers”). File > Project Settings > Headphone 3D Audio makes the game start with it on; it is unticked in a new project.

Tuning

The measurements come from one standard head, and players’ ears and headphones differ. Three settings let a player adjust it; all of them can change while sounds play, so they can sit on sliders in an options menu.

// In an options menu:
Audio.Hrtf = headphones;
Audio.HrtfStrength = strengthSlider;      // 0..1
Audio.HrtfHeadSize = headSizeSlider;      // 0.5..1.5
Audio.HrtfTreble = trebleSlider;          // -12..12

Values outside a range are clamped. The settings are not saved for you; store them with the rest of your options and set them again at start-up. A good menu plays a sound circling the player while they adjust.

The measurements are the MIT Media Lab’s KEMAR set (Gardner and Martin, 1994), which is built into the engine; nothing is loaded from disk. A sound moving round the listener glides from one measured direction to the next.

Volumes

Audio.MasterVolume = 0.8f;
Audio.SetGroupVolume(.Music, 0.5f);
Audio.SetGroupVolume(.Effects, 1.0f);

Every source plays into its group, and every group into the master, so an options menu is three sliders.

Reverb zones

Add an Audio Reverb Zone where there should be an echo: a hall, a cave, a tunnel. While the listener is inside, spatial sounds gain reverb, fading in across the zone’s edge.

Subtitles

A sound with a subtitle file beside it (Voice/intro.ogg and Voice/intro.srt) shows its lines while it plays, and an Audio Source’s Subtitles box turns that off. Conversations and the subtitle settings are on the Dialogue page (docs/dialogue.md).

Multiplayer

Play, Stop and PlayOneShot on an Audio Source are shared like animation: called on the server, or by the player who controls the object, everyone hears it. Called on a client for an object it does not control, only that client hears it. Audio.PlayOneShot is always local.

How good is it

Each source’s mixing only does what is in use. The low-pass filter is in its path only while it cuts something (below 20 kHz), and the reverb send only while it is above 0. A sound at pitch 1 with no Doppler, recorded at the mixer’s rate, skips resampling. HRTF off means plain panning. 100 looping positioned sources: 1.4 ms to 0.45 ms of mixing per 1,024 frames. HRTF costs about 0.11 ms per source per 1,024 frames.

What you get from a plain build: stereo panning, optional HRTF for headphones, distance attenuation with three rolloff curves, Doppler, directional cones, per-source low-pass used for occlusion and air absorption, a stereo reverb, six mixer groups, streaming, and any number of overlapping one-off sounds.

What it is not:

• One head for everyone. The HRTF is measured on a standard dummy head. It works well for most people, but ears differ, and a few listeners find front and back harder to tell apart than others do. It covers directions from 40 degrees below the horizon upward; sounds further below use the lowest measurement.
• No room in the HRTF. Sounds are placed in direction and distance, but walls only matter through occlusion and reverb zones; there are no reflections computed from the level’s shape.
• Occlusion is one ray from the listener to the source, ten times a second, eased between answers. A sound is either blocked or not; there is no partial blocking by thin objects and no sound bending round corners.
• Reverb comes from where the listener stands, not from the room the source is in, and one zone applies at a time.
• A dedicated server (no window) plays nothing.

CPU tests listen to the engine’s own mix with no sound card and measure it: left and right, distance, each rolloff, HRTF (off by default, the near ear louder and about 0.7 ms earlier, front, back, above and below each coloured differently, strength, head size and treble, no clicks as a sound circles or the settings are dragged), volumes and groups, looping, one-off sounds, a missing file, occlusion by a wall, air absorption, reverb tails, the camera as listener, and sounds shared over a network session.