Game Lab · Music & sound

Game audio with
a real score

Most generated game audio is a loop and some beeps. Every piece here is a score: multi-section themes with fills and B-sections, layers that react to play, and effects built from layered real instruments — playable in the browser, judged until clean.

Try the AI Game Builder

Form, not loops

A single loop repeating unchanged for a session is a failed cycle by definition. Every score has real musical form — theme, B-section, fills, variation — mapped bar by bar, so ten minutes of play never sounds like one minute repeated ten times.

Form map — "Harbor Lights"
A ×2
fill
B
A’
fill
92 BPM · 6/8A’ revoices the themeno bar repeats byte-identical

A score that plays the game with you

At least three intensity layers enter and exit musically — bar-aligned, never mid-phrase — as play heats and cools, with ducking under impacts. The gallery piece proves it: drag the intensity slider and hear the arrangement respond.

State layersintensity: 2 of 3
Base — felt piano + upright bassalways on
Mid — brushed kit + marimbaentered at bar 17, on the downbeat
Peak — strings + brass swellswaiting for intensity 3

The most-heard sounds get the most craft

A pick-up you hear four hundred times deserves more than one marimba note. Every effect is layered from ≥2 real instrument voices with pitch and velocity variation per trigger — the exact class of cheap-SFX defect that once blocked real customer builds.

SFX kit — coin pickup3 voices
Celesta strikeroot note, velocity 0.7–1.0 randomized
Glockenspiel +12+0–30 cents drift per trigger
Felt-piano thumplow body so it reads on phone speakers
400 triggers per session — and no two identical.

Judged by the same judge that gates real games

The platform’s audio verifier — the one that blocks customer builds on composition and SFX-design defects — is self-applied here at its strictest reading, plus numeric checks: section boundaries present, layer counts, spectral variety between sections.

Rubric self-scoreClean pass
Compositionform present, A’ revoiced, fills at turns
State layering3 layers, bar-aligned entries, ducking
SFX designmost-frequent sounds: 3 voices, varied
Numeric checksspectral variety A vs B: distinct

How a piece gets made

01

Brief

Genre, tempo, mood, instrumentation palette — committed first.

02

Compose

Theme, B-section, fills, variation — from 84 sampled instruments.

03

Layer + design

State-reactive arrangement; layered, varied SFX.

04

Judge until clean

The platform’s own audio rubric at its strictest, plus numbers.

05

Ship playable

An interactive player: score, intensity slider, every SFX.

Pieces

Each card is a piece produced by the loop above — newest first.

Questions, answered

Why sampled instruments instead of synthesis?

Platform decree, born from shipped defects: bare-oscillator audio reads as cheap instantly. Every sound — music and effects — comes from an 84-instrument sampled library of real instruments, in every genre.

What does "state-reactive" mean?

The score is arranged in layers tied to game intensity: calm play gets the base arrangement, action brings layers in on the bar, impacts duck the bed. Each gallery piece has an intensity slider so you can hear it respond.

How is a piece judged?

By the same audio verifier that gates customer games — composition form, layering, SFX craft — self-applied at its strictest reading, plus numeric checks on section boundaries and spectral variety. Findings iterate until a full pass is clean.

Does my game get audio like this?

Yes — every game the builder ships composes its score and sound kit from the same sampled library under the same standards this lab hardens.

Build a game that sounds like this

The builder composes every game’s score and sound kit from real sampled instruments.

Try the AI Game Builder