version: "1.0.0" name: psl1ght-audio description: >- Audio on PS3 via MikMod. Use when adding music or sound effects to PS3 homebrew — MikMod plays tracker modules / PCM samples (NOT OGG/MP3); synthesize SFX in code or load samples from memory via an MREADER; hand-built WAVs must be little-endian (the PPU is big-endian); loop long music as a sample; and init defensively because a bad audio init can hang the console.

Audio (MikMod)

MikMod plays tracker modules and samples (PCM) — it cannot decode OGG/MP3. Drive it with MikMod_Update() once per frame and reserve voices with MikMod_SetNumVoices(music, sfx) (sample playback needs sfx voices reserved).

raylib-on-PS3 has no working audio device. On the RSXGL/raylib stack, raylib's

InitAudioDevice/LoadSound/PlaySound aren't linkable (undefined references — even from

dead code). MikMod, below, is the audio path for raylib ports too: embed the game's .wavs

and load them with Sample_LoadGeneric via an MREADER (see Load audio from memory).

Synthesize SFX in code — no assets needed

You can generate sound effects procedurally: build a PCM waveform with math (a decaying sine for a "thud", a falling sweep for a "pew", noise for an explosion), wrap it as a little-endian WAV in memory, and load it as a MikMod sample. Self-contained (no asset files), great for prototyping, placeholders, or a stylized retro feel.

#define SR 22050
short pcm[SR/2];
int n = SR * 8 / 100;                       /* 80 ms */
for (int i = 0; i < n; i++) {
    float t = (float)i / SR, env = expf(-t * 22.0f);     /* fast decay */
    float s = sinf(2*M_PI*150.0f*t) * 0.7f + noisef() * 0.3f;
    pcm[i] = (short)(env * s * 22000.0f);
}
/* -> make_wav(pcm, n) into a byte buffer, then Sample_LoadGeneric (see below) */

noisef() is a tiny LCG mapped to −1..1. Vary frequency/decay/noise mix per effect. (Note: if you want to avoid the libm dependency for sinf/expf, a square wave via a manual phase wrap + linear envelope sounds fine and needs no math lib.)

Load audio from memory via an MREADER

MikMod's loaders (Sample_LoadGeneric, Player_LoadGeneric) take an `MREADER`, not a raw pointer. Implement a ~5-function in-memory reader over an embedded buffer (from bin2o) and pass &reader.core:

typedef struct { MREADER core; const unsigned char *data; long size, pos; } MemReader;
/* Get -> byte or EOF; Read -> 1 on FULL read else 0; Seek -> 0 on success (non-zero
 * error); Tell -> pos; Eof -> pos>=size.  NOTE the inverted Read vs Seek conventions. */
SAMPLE *s = Sample_LoadGeneric(&mr.core);   /* NULL on failure */

Embed the .wav with bin2o by naming the file *.bin in data/ (the Makefile's bin2o rule keys on extension; MikMod parses the RIFF header, not the filename).

Convert real assets to a clean canonical WAV first (strips odd chunks that loaders trip on), e.g. with ffmpeg:

ffmpeg -i in.ogg -ac 1 -ar 22050 -c:a pcm_s16le -map_metadata -1 -f wav data/music.bin

Hand-built WAV bytes must be little-endian (PPU is big-endian) ⚠️

WAV is little-endian by spec. If you build a WAV yourself (e.g. for synthesized SFX), write the header fields and the int16 PCM as explicit LE bytes — a naive u32/s16 store is big-endian on the PPU and MikMod will misread it. (Real .wav files from ffmpeg are already LE, so this only bites hand-built buffers.)

static void put_u16le(unsigned char *b, unsigned v){ b[0]=v; b[1]=v>>8; }
static void put_u32le(unsigned char *b, unsigned v){ b[0]=v; b[1]=v>>8; b[2]=v>>16; b[3]=v>>24; }

Loop long music as a sample (the OGG/MP3 way)

Since MikMod can't play OGG/MP3 and a module would mean re-authoring, decode the track to PCM WAV on the host (ffmpeg) — or synthesize a melody — and play it as one long looping sample: set SF_LOOP + loopend, and play with SFX_CRITICAL so transient SFX voices can't steal it.

music->flags |= SF_LOOP; music->loopstart = 0; music->loopend = music->length;
SBYTE v = Sample_Play(music, 0, SFX_CRITICAL);
if (v >= 0) Voice_SetVolume(v, 150);        /* duck under SFX */

Init audio defensively — it can hang the console ⚠️

A bad audio init can freeze a PS3 (double-init is a known hang), and you can't hear it on the dev host. Guard every step; on any failure set an audio_ok = 0 flag and make all calls no-ops (degrade to silence). Isolate audio in its own file so it's easy to disable.