Sound Design Deep Dive: Synthesis and Middleware

Volume 3 covered game audio concepts: SFX layering, pitch variation, adaptive music, UI audio. This article goes deeper into the craft: how sounds are synthesised, how impact sounds are architecturally constructed, and how professional studios use middleware like FMOD to build adaptive audio systems that feel alive.

Synthesis Fundamentals

Most game sounds are not pure recordings -- they are synthesised or heavily processed recordings. Understanding synthesis allows you to communicate more precisely with audio designers and to build basic effects yourself when budget or time constraints require it.

Subtractive synthesis starts with a harmonically rich waveform (sawtooth, square, noise) and removes frequencies with filters. It is the most common approach for game SFX: explosions, weapons, impacts. Additive synthesis builds complex timbres by layering sine waves at different frequencies and amplitudes. Better for magical effects, UI tones, and digital sounds. FM synthesis uses one oscillator to modulate another's frequency, producing metallic, bell-like, or electric tones efficiently -- ideal for retro-style SFX or mechanical sounds. Granular synthesis slices audio into tiny grains and rearranges or pitch-shifts them. Powerful for environmental ambience and voice distortion.

In practice, most professional game audio uses sampling plus layering: record real-world sound sources (metal striking metal, cloth tearing, water splashing), then layer, pitch-shift, filter, and process them heavily. The gunshot in Call of Duty bears little resemblance to the sound a gun actually makes -- it has been processed to feel powerful, not to sound realistic.

Impact Sound Architecture

A well-designed impact sound is not a single sample. It is three distinct layers playing simultaneously, each carrying different perceptual information:

The transient layer (0-20ms) is the sharp crack or thwack. High frequency, fast attack, near-instant decay. This is what the brain registers as impact happened. It is the first layer to cut through background noise and the first to register consciously. Without it, impacts sound soft and cushioned.

The body layer (20-200ms) carries the weight and material character of the impact. A metal hit has metallic ring; flesh has a wet thud; stone has a dull crack. Mid-frequency dominant. This layer is what distinguishes a sword hit from a fist hit from a hammer hit -- all three have similar transients, but completely different body layers.

The sub layer (0-500ms) is a low-frequency rumble below 80Hz that the player feels through speakers or headphones. It creates the visceral gut-punch quality that separates a satisfying hit from a flat one. The sub layer is not always audible through laptop speakers -- but on any reasonable speaker or headphone setup, it is the layer that creates physical sensation.

Remove any layer and the sound loses a dimension: transient-only sounds sterile and thin; body-only sounds muffled and weak; sub-only is meaningless boom. The interplay of all three at the right relative amplitudes is what makes the Doom 2016 Super Shotgun sound the way it does. When a sound is not working, identify which layer is missing or imbalanced before reaching for a different sample entirely.

FMOD for Adaptive Audio

FMOD Studio is the most widely-used game audio middleware. It sits between your game engine and the audio playback hardware, enabling logic-driven, parameter-responsive audio that is impossible to implement with raw AudioSource calls.

Events are the fundamental unit of FMOD audio. Each sound effect is an Event. Events contain multi-layer logic, randomisation, and parameter-driven variation in a visual editor. A single gunshot Event might contain 6 sample variants, pitch randomisation, a low-pass filter that responds to distance, and a reverb send that responds to room size -- all without any per-instance code.

Parameters are float values sent from the game to FMOD (player_health, enemy_proximity, combat_intensity). These drive automated audio changes. As combat_intensity rises, music layers stack in. As player_health drops, a heartbeat event fades in. All the logic lives in FMOD; the game sends one number.

Horizontal resequencing and vertical remixing: FMOD enables music to switch between looping regions (combat, exploration, stealth) with musically seamless transitions based on game parameters. Vertical remixing adds or removes instrument layers as intensity changes -- the combat theme might add a percussion layer as the fight escalates, then remove it as enemies die. Celeste uses FMOD for its entire dynamic music system.

The No Two Shots Alike Principle

In Hades, no two sword strikes sound identical. The audio team created pools of 6-8 variants per weapon per action type, with pitch and filter randomisation on top. The result is thousands of effective combinations. Players will never consciously notice this -- but they will notice its absence as audio fatigue after 20 minutes of combat.

The minimum viable version: for your highest-frequency sound (the most common gameplay action), create 4-6 variants and add plus/minus 5-10% pitch randomisation. This alone eliminates the most common audio fatigue pattern. Scale up to larger pools as budget allows, targeting 6-8 variants for all primary combat sounds and 3-4 variants for secondary events.

Wwise vs FMOD

FMOD Studio is the better choice for most indie and mid-size teams: moderate learning curve, free under certain revenue thresholds, strong Unity and Unreal integration, well-documented community resources. Wwise (Audiokinetic) is better suited to AA and AAA productions with complex bus mixing requirements and dedicated audio teams. Both offer free tiers. For most indie games, FMOD will do everything needed and is faster to learn.