Audio Primitives

Symphonia provides a comprehensive set of primitives for working with decoded audio data. These primitives handle different sample formats, provide efficient planar audio storage, and enable safe conversion between formats.

Core Audio Types

Symphonia’s audio system is built around several key types:

AudioBuffer and AudioBufferRef

AudioBuffer<S>

AudioBuffer<S> (symphonia-core/src/audio.rs:283) is the fundamental audio container. It stores multi-channel audio in a planar format.

pub struct AudioBuffer<S: Sample> {
    buf: Vec<S>,           // All channel data
    spec: SignalSpec,      // Signal characteristics
    n_frames: usize,       // Number of frames written
    n_capacity: usize,     // Maximum frames
}

Planar vs Interleaved Layout

Planar (AudioBuffer)
Interleaved (SampleBuffer)

Samples are stored channel-by-channel:

[L L L L L L L L] [R R R R R R R R]
 ^^^^^^^^^^^^^^     ^^^^^^^^^^^^^^
 Left channel       Right channel

Benefits:

Cache-friendly for DSP operations
Easier to apply per-channel processing
Native format for most decoders

Samples are stored frame-by-frame:

[L R L R L R L R L R L R L R L R]
 ^^^ ^^^ ^^^ ^^^ ^^^ ^^^ ^^^ ^^^
 Frames (stereo pairs)

Benefits:

Expected by most audio APIs (PortAudio, CPAL, etc.)
Easier to write to audio hardware
Natural for sample-aligned operations

AudioBufferRef

AudioBufferRef (symphonia-core/src/audio.rs:419) is an any-type wrapper returned by decoders. It hides the specific sample format using an enum:

pub enum AudioBufferRef<'a> {
    U8(Cow<'a, AudioBuffer<u8>>),
    U16(Cow<'a, AudioBuffer<u16>>),
    U24(Cow<'a, AudioBuffer<u24>>),
    U32(Cow<'a, AudioBuffer<u32>>),
    S8(Cow<'a, AudioBuffer<i8>>),
    S16(Cow<'a, AudioBuffer<i16>>),
    S24(Cow<'a, AudioBuffer<i24>>),
    S32(Cow<'a, AudioBuffer<i32>>),
    F32(Cow<'a, AudioBuffer<f32>>),
    F64(Cow<'a, AudioBuffer<f64>>),
}

Decoders return AudioBufferRef because different codecs produce different sample formats. Your application must handle this polymorphism.

Working with AudioBufferRef

let decoded = decoder.decode(&packet)?;

match decoded {
    AudioBufferRef::F32(buf) => {
        // Process f32 samples
        for &sample in buf.chan(0) {
            println!("Sample: {}", sample);
        }
    }
    AudioBufferRef::S16(buf) => {
        // Process i16 samples
    }
    _ => {
        // Handle other formats
    }
}

Sample Formats

Symphonia supports all common audio sample formats via the Sample trait.

Supported Sample Types

Integer
Floating Point
Special

Type	Bits	Range	Use Case
`u8`	8	0 to 255	Legacy formats, WAV
`i8`	8	-128 to 127	Telephony
`u16`	16	0 to 65535	-
`i16`	16	-32768 to 32767	CD audio, most hardware
`u24`	24	0 to 16777215	High-res audio
`i24`	24	-8388608 to 8388607	Studio recording
`u32`	32	0 to 4294967295	-
`i32`	32	-2147483648 to 2147483647	Studio processing

Type	Bits	Range	Use Case
`f32`	32	-1.0 to 1.0	DSP, modern DAWs
`f64`	64	-1.0 to 1.0	Scientific, high precision

Floating-point formats use normalized range [-1.0, 1.0] where 0.0 is silence.

u24 and i24 are custom 24-bit integer types:

use symphonia::core::sample::{u24, i24};

let sample = i24::from_i32(8388607);
let as_i32 = sample.to_i32();
let bytes = sample.to_ne_bytes();  // [u8; 3]

These types are used by some formats (FLAC, ALAC, high-res WAV) and are efficiently packed.

Sample Trait

All sample types implement the Sample trait, which provides:

pub trait Sample: Copy + Clone + Send + Sync + ... {
    /// The equilibrium value (silence)
    const MID: Self;
    
    // Conversion traits are implemented separately
}

The MID constant represents silence:

Unsigned: mid-point (128 for u8, 32768 for u16)
Signed: zero (0)
Float: zero (0.0)

Sample Conversions

Symphonia provides automatic sample format conversion:

use symphonia::core::conv::IntoSample;

let i16_sample: i16 = 16384;
let f32_sample: f32 = i16_sample.into_sample();  // ≈ 0.5

let f32_sample: f32 = 0.5;
let i16_sample: i16 = f32_sample.into_sample();  // 16384

SampleBuffer

SampleBuffer<S> (symphonia-core/src/audio.rs:717) converts AudioBufferRef to a typed, interleaved sample buffer.

Creating and Using SampleBuffer

use symphonia::core::audio::SampleBuffer;

// Create buffer for decoded audio
let mut sample_buf = SampleBuffer::<f32>::new(
    decoded.capacity() as u64,
    *decoded.spec(),
);

// Copy interleaved (LRLRLR...)
sample_buf.copy_interleaved_ref(decoded);

// Access samples
let samples: &[f32] = sample_buf.samples();
let len = sample_buf.len();  // Total sample count

Always recreate SampleBuffer if you receive Error::ResetRequired from the decoder, as the signal spec may have changed.

RawSampleBuffer

RawSampleBuffer<S> (symphonia-core/src/audio.rs:997) is like SampleBuffer, but provides samples as raw bytes.

When to Use RawSampleBuffer

Use RawSampleBuffer when:

Interfacing with C APIs
Writing to file formats
Sending over network sockets
Working with FFI boundaries

use symphonia::core::audio::RawSampleBuffer;

let mut raw_buf = RawSampleBuffer::<f32>::new(
    decoded.capacity() as u64,
    *decoded.spec(),
);

// Copy as interleaved bytes
raw_buf.copy_interleaved_ref(decoded);

// Get as byte slice
let bytes: &[u8] = raw_buf.as_bytes();

SignalSpec

SignalSpec (symphonia-core/src/audio.rs:164) describes the characteristics of an audio signal.

pub struct SignalSpec {
    pub rate: u32,         // Sample rate in Hz
    pub channels: Channels, // Channel configuration
}

Creating SignalSpec

use symphonia::core::audio::{SignalSpec, Channels};

// Stereo at 44.1kHz
let spec = SignalSpec::new(
    44100,
    Channels::FRONT_LEFT | Channels::FRONT_RIGHT
);

Channels and Layouts

Channels Bitmask

Channels (symphonia-core/src/audio.rs:36) is a bitmask representing channel configuration:

use symphonia::core::audio::Channels;

// Mono
let mono = Channels::FRONT_LEFT;

// Stereo
let stereo = Channels::FRONT_LEFT | Channels::FRONT_RIGHT;

// 2.1 (Stereo + LFE)
let stereo_lfe = Channels::FRONT_LEFT 
               | Channels::FRONT_RIGHT 
               | Channels::LFE1;

// 5.1 Surround
let surround = Channels::FRONT_LEFT
             | Channels::FRONT_RIGHT
             | Channels::FRONT_CENTRE
             | Channels::REAR_LEFT
             | Channels::REAR_RIGHT
             | Channels::LFE1;

Channel Layouts

Pre-defined layouts for common configurations:

pub enum Layout {
    Mono,           // Single center channel
    Stereo,         // Left + Right
    TwoPointOne,    // Left + Right + LFE
    FivePointOne,   // Front L/R/C + Rear L/R + LFE
}

// Convert to Channels
let channels = Layout::Stereo.into_channels();

Signal Trait

The Signal trait (symphonia-core/src/audio.rs:501) provides methods for manipulating audio buffers:

use symphonia::core::audio::Signal;

// Render silence
buffer.render_silence(Some(1024));

// Render with custom function
buffer.render(Some(1024), |planes, frame_idx| {
    for plane in planes.planes() {
        plane[frame_idx] = 0.0;  // Generate samples
    }
    Ok(())
})?;

Best Practices

Always Reuse Buffers

Create SampleBuffer and RawSampleBuffer once and reuse them:

// Good
let mut sample_buf = SampleBuffer::<f32>::new(capacity, spec);
loop {
    let decoded = decoder.decode(&packet)?;
    sample_buf.copy_interleaved_ref(decoded);  // Reuse
}

// Bad - allocates every iteration
loop {
    let decoded = decoder.decode(&packet)?;
    let mut sample_buf = SampleBuffer::<f32>::new(capacity, spec);
    sample_buf.copy_interleaved_ref(decoded);
}

Prefer f32 for Processing

Use f32 samples for DSP operations:

Normalized range [-1.0, 1.0]
No overflow/underflow concerns
Native format for most audio libraries
Good precision for audio work

Handle Format Changes

Be prepared for Error::ResetRequired:

match decoder.decode(&packet) {
    Ok(decoded) => { /* process */ }
    Err(Error::ResetRequired) => {
        // Recreate buffers with new spec
        decoder.reset();
        // Re-examine codec parameters
    }
    Err(e) => return Err(e),
}

Use Direct Access When Possible

If you only need one channel or specific processing, access channels directly:

// More efficient than copying to SampleBuffer
match decoded {
    AudioBufferRef::F32(buf) => {
        let left = buf.chan(0);
        let right = buf.chan(1);
        // Process directly
    }
    _ => { /* handle other formats */ }
}

Next Steps

Media Sources

Learn how Symphonia reads audio data from different sources

Decoding Audio

Complete guide to decoding audio files

Processing Audio

Apply effects and transformations to decoded audio

Exporting Audio

Convert and export audio to different formats

Get Started

Core Concepts

Guides

Resources

Core Audio Types

AudioBuffer and AudioBufferRef

AudioBuffer<S>

Planar vs Interleaved Layout

AudioBufferRef

Working with AudioBufferRef

Sample Formats

Supported Sample Types

Sample Trait

Sample Conversions

SampleBuffer

Creating and Using SampleBuffer

RawSampleBuffer

When to Use RawSampleBuffer

SignalSpec

Creating SignalSpec

Channels and Layouts

Channels Bitmask

Channel Layouts

Signal Trait

Best Practices

Next Steps

Media Sources

Decoding Audio

Processing Audio

Exporting Audio

Build docs developers (and LLMs) love

Get Started

Core Concepts

Guides

Resources

​Core Audio Types

​AudioBuffer and AudioBufferRef

​AudioBuffer<S>

​Planar vs Interleaved Layout

​AudioBufferRef

​Working with AudioBufferRef

​Sample Formats

​Supported Sample Types

​Sample Trait

​Sample Conversions

​SampleBuffer

​Creating and Using SampleBuffer

​RawSampleBuffer

​When to Use RawSampleBuffer

​SignalSpec

​Creating SignalSpec

​Channels and Layouts

​Channels Bitmask

​Channel Layouts

​Signal Trait

​Best Practices

​Next Steps

Media Sources

Decoding Audio

Processing Audio

Exporting Audio

Build docs developers (and LLMs) love

Core Audio Types

AudioBuffer and AudioBufferRef

AudioBuffer<S>

Planar vs Interleaved Layout

AudioBufferRef

Working with AudioBufferRef

Sample Formats

Supported Sample Types

Sample Trait

Sample Conversions

SampleBuffer

Creating and Using SampleBuffer

RawSampleBuffer

When to Use RawSampleBuffer

SignalSpec

Creating SignalSpec

Channels and Layouts

Channels Bitmask

Channel Layouts

Signal Trait

Best Practices

Next Steps