diff --git a/Cargo.toml b/Cargo.toml
index aca94e4..b59c19d 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -16,6 +16,7 @@ members = [
     "examples/shadow_demo",
     "examples/ibl_demo",
     "crates/voltex_physics",
+    "crates/voltex_audio",
 ]
 
 [workspace.dependencies]
@@ -25,6 +26,7 @@ voltex_renderer = { path = "crates/voltex_renderer" }
 voltex_ecs = { path = "crates/voltex_ecs" }
 voltex_asset = { path = "crates/voltex_asset" }
 voltex_physics = { path = "crates/voltex_physics" }
+voltex_audio = { path = "crates/voltex_audio" }
 wgpu = "28.0"
 winit = "0.30"
 bytemuck = { version = "1", features = ["derive"] }
diff --git a/crates/voltex_audio/Cargo.toml b/crates/voltex_audio/Cargo.toml
new file mode 100644
index 0000000..e1ac68f
--- /dev/null
+++ b/crates/voltex_audio/Cargo.toml
@@ -0,0 +1,6 @@
+[package]
+name = "voltex_audio"
+version = "0.1.0"
+edition = "2021"
+
+[dependencies]
diff --git a/crates/voltex_audio/src/audio_clip.rs b/crates/voltex_audio/src/audio_clip.rs
new file mode 100644
index 0000000..b5dddd4
--- /dev/null
+++ b/crates/voltex_audio/src/audio_clip.rs
@@ -0,0 +1,55 @@
+/// A loaded audio clip stored as normalized f32 PCM samples.
+#[derive(Debug)]
+pub struct AudioClip {
+    /// Interleaved PCM samples in the range [-1.0, 1.0].
+    pub samples: Vec<f32>,
+    /// Number of samples per second (e.g. 44100).
+    pub sample_rate: u32,
+    /// Number of audio channels (1 = mono, 2 = stereo).
+    pub channels: u16,
+}
+
+impl AudioClip {
+    /// Create a new AudioClip from raw f32 samples.
+    pub fn new(samples: Vec<f32>, sample_rate: u32, channels: u16) -> Self {
+        Self {
+            samples,
+            sample_rate,
+            channels,
+        }
+    }
+
+    /// Total number of multi-channel frames.
+    /// A frame contains one sample per channel.
+    pub fn frame_count(&self) -> usize {
+        self.samples.len() / self.channels as usize
+    }
+
+    /// Duration in seconds.
+    pub fn duration(&self) -> f32 {
+        self.frame_count() as f32 / self.sample_rate as f32
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn mono_clip() {
+        // 44100 Hz, 1 channel, 1 second of silence
+        let samples = vec![0.0f32; 44100];
+        let clip = AudioClip::new(samples, 44100, 1);
+        assert_eq!(clip.frame_count(), 44100);
+        assert!((clip.duration() - 1.0).abs() < 1e-6);
+    }
+
+    #[test]
+    fn stereo_clip() {
+        // 44100 Hz, 2 channels, 0.5 seconds
+        let samples = vec![0.0f32; 44100]; // 44100 interleaved samples = 22050 frames
+        let clip = AudioClip::new(samples, 44100, 2);
+        assert_eq!(clip.frame_count(), 22050);
+        assert!((clip.duration() - 0.5).abs() < 1e-6);
+    }
+}
diff --git a/crates/voltex_audio/src/lib.rs b/crates/voltex_audio/src/lib.rs
new file mode 100644
index 0000000..91abd33
--- /dev/null
+++ b/crates/voltex_audio/src/lib.rs
@@ -0,0 +1,7 @@
+pub mod audio_clip;
+pub mod wav;
+pub mod mixing;
+
+pub use audio_clip::AudioClip;
+pub use wav::{parse_wav, generate_wav_bytes};
+pub use mixing::{PlayingSound, mix_sounds};