#Game sound effects (WAV)

Available in npm 0.2.0. Check release status for the current package version and validation scope.

Use Format.Wav and components from @delino/react-forge/sfx to compose procedural game effects. Export a WAV buffer or file for your game engine. Generation is offline and does not need recordings, audio devices, fonts or a conversion tool.

#A zombie-game gunshot

Save this as gunshot.tsx:

import React from "react";
import { createSession, Format } from "@delino/react-forge";
import { Sound, Noise, Tone, SampleRate } from "@delino/react-forge/sfx";

export default async function task() {
  const session = createSession(Format.Wav);
  await session.render(
    <Sound duration={0.75} sampleRate={SampleRate.Game}>
      <Noise duration={0.095} seed={815} gain={0.85}
        attack={0.0003} release={0.025} decay={0.017}
        highPass={1100} lowPass={15000} />
      <Noise start={0.002} duration={0.21} seed={816} gain={1.4}
        attack={0.001} release={0.06} decay={0.044} lowPass={1800} />
      <Tone duration={0.17} frequency={155} endFrequency={48}
        gain={0.8} attack={0.0008} release={0.05} decay={0.032} />
      <Noise start={0.075} duration={0.035} seed={817} gain={0.14}
        attack={0.0005} release={0.01} decay={0.009}
        highPass={2400} lowPass={11000} />
      <Noise start={0.022} duration={0.58} seed={818} gain={0.24}
        attack={0.006} release={0.18} decay={0.13}
        highPass={160} lowPass={3200} />
    </Sound>,
  );
  return session;
}

react-forge run gunshot.tsx --output zombie-gunshot.wav --json

The layers provide a sharp crack, low body, falling tone, mechanical click and short room tail. Changing the noise seeds gives variations. This is original procedural game audio, not a recording or a claim of realistic weapon acoustics. Exported audio is 16-bit PCM WAV. Listen in your target game's mix to choose the appropriate gain. Files are only generated on explicit export; nothing plays automatically.

#Components and parameters

Layer parameters are required duration, optional start, gain, pan, attack, release and decay. Times are seconds and gain is linear. Each layer must fit entirely inside its Sound. Noise and Tone are leaves; use ordinary React components and fragments to reuse or repeat layers.

Defaults are SampleRate.Game (48000 Hz), Channels.Mono, unity gain and start zero. SampleRate.Cd selects 44100 Hz. Channels.Stereo enables equal-power pan from -1 (left) to 1 (right); mono requires pan zero. Gains range from zero to four. Tone defaults to Waveform.Sine; Waveform.Triangle is also available. Tone frequency sweeps linearly to endFrequency when supplied.

Attack and release provide linear fades. They default to 1 ms and 10 ms, each capped at one quarter of the layer and floored at one sample. Each must be at least one sample and together must fit the layer. Optional decay is an exponential amplitude time constant from one sample to 30 seconds. The first and last sample of each layer are zero. Peaks above 0.95 are attenuated together without boosting quiet sounds or changing stereo balance.

Seeds are integers from 1 through 4294967295. Noise filters and tone frequencies range from 20 Hz to one Hz below half the sample rate. When both filters are used, highPass must be below lowPass. Repeatable seeds reproduce the same sound on the same runtime and host; byte identity across platforms is not guaranteed.

#Sessions, MCP and limits

Use the common session APIs, CLI or MCP tools. MCP execution accepts the same TSX task; inspect its session and call react_forge_export with an output ending in .wav. measure returns timeline geometry: x is the sample-rounded start and width is duration in seconds, with y zero and height one. A changed revision invalidates an earlier measurement.

Sounds last from two samples to 30 seconds, with one to 256 layers. The sum of layer durations in samples cannot exceed 16 million. capabilities exposes these limits. Invalid latest renders fail export; cancellation and failed exports preserve existing destinations. Existing files require explicit overwrite. Library callers should dispose sessions in finally.

SFX supports generation only. Audio import, recorded samples, MP3/OGG, MIDI, microphone capture and live playback are not supported. WAV sessions reject image and font registration. The 0.2.0 release workflow exercised installed consumers on all six supported native hosts. No game-engine listening evaluation is claimed; see validation scope.

For animated pixel artwork instead of audio, see Sprite.