NAudio

Audio and MIDI library for .NET

6,140 stars C#
RAW Doc

File: README.md

NAudio

NAudio is an open source .NET audio library written by Mark Heath

📖 Documentation site — tutorials and the full API reference.

NAudio 3

NAudio 3 is a major release. The headlines:

  • The single NAudio assembly is split into focused packages — take only what you need. The NAudio meta-package still pulls the Windows stack together, so existing consumers see no change.
  • The core is cross-platform and Native-AOT compatible. NAudio.Core, NAudio.Midi, NAudio.Effects, NAudio.Sampler and NAudio.SoundFile run on Windows, Linux and macOS.
  • Minimum target framework is net9.0 — legacy .NET Framework and .NET Standard 2.0 support is dropped.
  • New subsystems: an audio effects framework, a software sampler, VST 3 hosting, ALSA playback/capture on Linux, and cross-platform file I/O via libsndfile.
  • Modernised WASAPI and ASIO — the new WasapiPlayer / WasapiRecorder and AsioDevice APIs.

Upgrading from NAudio 2? Start with Migrating from NAudio 2 to NAudio 3. The full list of changes is in RELEASE_NOTES.md.

Packages

Installing the NAudio meta-package gets you the full Windows stack and is the right default. On a non-Windows target framework it resolves to the cross-platform pieces only. Reference the individual packages directly if you want a smaller surface.

Package Platform What it gives you
NAudio any Meta-package — the Windows stack plus AudioFileReader and Mp3FileReader
NAudio.Core cross-platform WaveStream / ISampleProvider model, WAV & AIFF I/O, mixing, resampling, DSP, NAudio.Effects, sequencing
NAudio.Midi cross-platform (+ WinRT MIDI on Windows) MIDI event model, Standard MIDI File reading & writing
NAudio.Wasapi Windows WASAPI playback, capture and loopback; Media Foundation codecs
NAudio.WinMM Windows WaveOut / WaveIn, classic MIDI I/O, ACM codecs, mixer controls
NAudio.Asio Windows Low-latency multichannel playback and capture through ASIO drivers
NAudio.Dmo Windows DMO effects, DMO MP3 decoder and resampler, DirectSoundOut
NAudio.WinForms Windows WinForms controls, and the window-callback WaveOutWindow / WaveInWindow
NAudio.Sampler cross-platform Polyphonic software sampler — SoundFont (.sf2), SFZ and single-sample instruments
NAudio.SoundFile cross-platform Read and write WAV/AIFF/FLAC/Ogg-Vorbis/Opus/MP3 via libsndfile
NAudio.Alsa Linux AlsaOut / AlsaIn playback and capture via libasound
NAudio.Vst3 Windows VST 3 plug-in hosting (preview) — effects and instruments
NAudio.Extras cross-platform (+ Windows extras) Opinionated helpers — playback engine, capture mixing, ID3 tags

NAudio.Core, NAudio.Midi, NAudio.Wasapi, NAudio.Dmo, NAudio.Sampler, NAudio.SoundFile and NAudio.Alsa are Native-AOT compatible. See the assembly layout plan for the reasoning behind the split.

Documentation

NAudio comes with several demo applications, which are the quickest way to see how the various features fit together: NAudioDemo (WinForms), NAudioWpfDemo, and the smaller NAudioConsoleTest, AudioFileInspector, MidiFileConverter and MixDiff tools. They have the advantage of being kept up to date, whilst some of the tutorials you will find on the internet refer to old versions of NAudio.

Features

  • Play back audio using a variety of APIs
    • WASAPI (WasapiPlayer, and the legacy WasapiOut)
    • WaveOut
    • ASIO
    • DirectSound
    • ALSA on Linux
  • Read audio from many standard file formats
    • WAV, AIFF and raw PCM
    • MP3 (using ACM, DMO or MFT)
    • G.711 mu-law and a-law
    • ADPCM, G.722, Opus (using Concentus)
    • WMA, AAC, MP4 and more with Media Foundation
    • FLAC, Ogg Vorbis, Opus and MP3 cross-platform with libsndfile
  • Convert between various forms of uncompressed audio
    • Change the number of channels — mono to stereo, stereo to mono, and arbitrary matrix routings
    • Modify bit depth (8, 16, 24, 32 integer or 32 bit IEEE float)
    • Resample audio using a choice of resampling algorithms
  • Encode audio using any ACM or Media Foundation codec installed on your computer
    • Create MP3s, AAC/MP4 audio and WMA files
    • Create WAV files containing G.711, ADPCM, G.722, etc.
    • Encode FLAC, Ogg Vorbis and Opus on any platform with NAudio.SoundFile
  • Mix and manipulate audio streams using a 32-bit floating point mixing engine
    • construct signal chains
    • examine sample levels for the purposes of metering or waveform rendering
    • pass blocks of samples through an FFT for metering or DSP
    • delay, loop, or fade audio in and out
  • Apply audio effects with the cross-platform NAudio.Effects framework
    • EQ and filtering, dynamics (compressor, limiter, gate, multiband), saturation and lo-fi
    • delay and modulation, reverb including FFT convolution, pitch shifting
    • click-free bypass, dry/wet mix and a parameter model for automation
  • Record audio using a variety of capture APIs
    • WASAPI (WasapiRecorder), including system audio and per-process loopback
    • WaveIn
    • ASIO
    • ALSA on Linux
  • Host VST 3 effects and instruments
  • Play SoundFont (.sf2) and SFZ instruments with the built-in software sampler
  • Work with soundcards
    • Enumerate devices
    • Access soundcard controls and metering information
    • Follow the default device automatically, and observe endpoint changes as events
  • Full MIDI event model
    • Read and write MIDI files
    • Respond to received MIDI events
    • Send MIDI events
    • Render a MIDI file to audio through the sampler or a hosted VST 3 instrument
  • An extensible programming model
    • All base classes easily inherited from for you to add your custom components

Tutorials

Upgrading

Playback

Working with Codecs

Working with audio files

Manipulating audio

Generating audio

Recording

Visualization

MIDI

Networking

Cross-platform and Linux

NAudio Training Courses

If you want to get up to speed as quickly as possible with NAudio programming, I recommend you watch these two Pluralsight courses. You will need to be a subscriber to access the content, but there is 10 hours of training material on NAudio, and it also will give you access to their vast training library on other programming topics.

To be successful developing applications that process digital audio, there are some key concepts that you need to understand. To help developers quickly get up to speed with what they need to know before trying to use NAudio, I have created the Digital Audio Fundamentals course, which covers sample rates, bit depths, file formats, codecs, decibels, clipping, aliasing, synthesis, visualisations, effects and much more. In particular, the fourth module on signal chains is vital background information if you are to be effective with NAudio.

Audio Programming with NAudio is a follow-on course which contains seven hours of training material covering all the major features of NAudio. It is highly recommended that you take this course if you intend to create an application with NAudio.

Please note that these courses were recorded against earlier versions of NAudio. The concepts all still apply, but some of the class names have changed — see Migrating from NAudio 2 to NAudio 3.

FAQ

What is NAudio?

NAudio is an open source audio API for .NET written in C# by Mark Heath, with contributions from many other developers. It is intended to provide a comprehensive set of useful utility classes from which you can construct your own audio application.

Why NAudio?

NAudio was created because the Framework Class Library that shipped with .NET 1.0 had no support for playing audio. The System.Media namespace introduced in .NET 2.0 provided a small amount of support, and the MediaElement in WPF and Silverlight took that a bit further. The vision behind NAudio is to provide a comprehensive set of audio related classes allowing easy development of utilities that play or record audio, or manipulate audio files in some way.

Does NAudio work on Linux and macOS?

Partly, and much more so in NAudio 3. NAudio.Core, NAudio.Midi, NAudio.Effects, NAudio.Sampler and NAudio.SoundFile are fully cross-platform, so signal chains, file I/O, DSP, effects and MIDI all work anywhere .NET runs. For output and capture, Linux has NAudio.Alsa; the WASAPI, WinMM, ASIO, DMO and WinForms packages remain Windows-only, and there is no macOS (CoreAudio) backend yet (although there is a work-in-progress implementation).

Which .NET versions are supported?

NAudio 3 requires net9.0 or later. If you need .NET Framework or .NET Standard 2.0, stay on NAudio 2.x.

Can I Use NAudio in my Project?

NAudio is licensed under the MIT license which means that you can use it in whatever project you like including commercial projects. Of course we would love it if you share any bug-fixes or enhancements you made to the original NAudio project files.

Is .NET Performance Good Enough for Audio?

While .NET cannot compete with unmanaged languages for very low latency audio work, it still performs better than many people would expect. On a fairly modest PC, you can quite easily mix multiple WAV files together, including pass them through various effects and codecs, play back glitch free with a latency of around 50ms.

How can I get help?

There are three main ways to get help. First, you can raise an issue here on GitHub. This is the best option when you've written some code and want to ask why it's not working as you expect. I attempt to answer all questions, but since this is a spare time project, occasionally I get behind.

You can also ask on StackOverflow and tag your question with naudio, if your question is a "how do I..." sort of question. This gives you a better chance of getting a quick answer. Please try to search first to see if your question has already been answered elsewhere.

Finally, I am occasionally able to offer paid support for situations where you need quick advice, bugfixes or new features. Please contact Mark Heath directly if you wish to pursue this option.

How do I submit a patch?

I welcome contributions to NAudio and have accepted many patches, but if you want your code to be included, please familiarise yourself with the following guidelines:

  • Your submission must be your own work, and able to be released under the MIT license.
  • You may use AI-assisted code generation, but ensure that your code is original and not copied from other sources. Note that I may choose to rewrite contributions rather than merge pull requests, so consider raising a feature request rather than a pull request first.
  • You will need to make sure your code conforms to the layout and naming conventions used elsewhere in NAudio.
  • Remember that there are many existing users of NAudio. A patch that changes the public interface is not likely to be accepted.
  • Try to write "clean code" - avoid long functions and long classes. Try to add a new feature by creating a new class rather than putting loads of extra code inside an existing one.
  • I don't usually accept contributions I can't test, so please write unit tests (using NUnit) if at all possible. If not, give a clear explanation of how your feature can be unit tested and provide test data if appropriate. Tell me what you did to test it yourself, including what operating systems and soundcards you used.
  • If you are adding a new feature, please consider writing a short tutorial on how to use it.
  • Unless your patch is a small bugfix, I will code review it and give you feedback. You will need to be willing to make the recommended changes before it can be integrated into the main code.
  • Patches should be provided using the Pull Request feature of GitHub.
  • Please also bear in mind that when you add a feature to NAudio, that feature will generate future support requests and bug reports. Are you willing to stick around on the forums and help out people using it?

File: Docs/AsioPlayback.md

Playback with ASIO

NAudio 3 introduces AsioDevice, a redesigned ASIO API that handles playback, recording, and duplex I/O through three explicit configuration modes. This article covers playback only — see AsioRecording and AsioDuplex for the other modes, and AsioMigration if you're moving from the legacy AsioOut class.

ASIO is the low-latency driver format supported by most professional Windows audio interfaces and many DAW applications. To use it you need a soundcard with an ASIO driver installed. If your hardware doesn't ship one, ASIO4ALL is a free WDM-to-ASIO shim that works with most consumer soundcards.

Open the device

Enumerate the installed ASIO drivers and open one by name:

c#
foreach (var name in AsioDevice.GetDriverNames())
    Console.WriteLine(name);

using var device = AsioDevice.Open("Focusrite USB ASIO");

AsioDevice implements IDisposable. Always wrap it in a using statement (or call Dispose explicitly) — the underlying COM driver doesn't release until you do.

Configure for playback

Pass an IWaveProvider (or wrap an ISampleProvider via .ToWaveProvider()) to InitPlayback:

c#
using var reader = new AudioFileReader("music.wav");

device.InitPlayback(new AsioPlaybackOptions
{
    Source = reader
});

The source's sample rate must be one the driver supports — device.IsSampleRateSupported(rate) answers that. The source's channel count must equal the number of output channels you select (defaults to a contiguous range starting at channel 0).

Select output channels

AsioPlaybackOptions.OutputChannels is an int[] of physical channel indices. Source channel n is routed to physical output OutputChannels[n]. The array can be non-contiguous — there's no ChannelOffset style restriction.

c#
// Stereo source → physical outputs 4 and 5 (zero-based).
device.InitPlayback(new AsioPlaybackOptions
{
    Source = reader,
    OutputChannels = [4, 5]
});

To send to every available output:

c#
OutputChannels = device.Capabilities.AllOutputChannels

device.Capabilities.NbOutputChannels tells you how many physical outputs the driver exposes; device.Capabilities.OutputChannelInfos[i].name gives a human-readable name for each.

See AsioChannelMapping for more channel-routing patterns.

Need to write each output channel independently — a different source per channel, rather than one interleaved IWaveProvider fanned out — for example to route separate channel pairs to separate speakers? Use duplex mode with no input channels.

Start and stop

c#
device.Start();
// ...
device.Stop();

Stop() raises the Stopped event on the captured SynchronizationContext (the thread you constructed the device on, typically the UI thread). The handler may safely call Dispose() — the device is fully off the ASIO callback thread by the time Stopped fires.

By default the device auto-stops when the source reaches end-of-stream. Set AutoStopOnEndOfStream = false in the options if you want the device to keep running on silent buffers after the source runs dry (e.g. so you can swap providers).

Handle errors and end-of-stream

c#
device.Stopped += (sender, e) =>
{
    if (e.Exception is not null)
        Console.WriteLine($"ASIO faulted: {e.Exception.Message}");
    else
        Console.WriteLine("Playback complete.");
};

Stopped fires exactly once per Start/Stop cycle, with e.Exception populated if the source threw or the driver reported an unrecoverable fault.

Recover from driver settings changes

If the user opens the driver's control panel and changes the sample rate (or any other setting), the driver fires a reset request. The recommended response:

c#
device.DriverResetRequest += (_, _) =>
{
    device.Stop();
    device.Reinitialize();
    device.Start();
};

Reinitialize() re-applies the most recent InitPlayback options against the (possibly changed) driver state — see AsioDriverReset for the full pattern.

Buffer size and latency

AsioPlaybackOptions.BufferSize accepts a frame count, or null to use the driver's preferred size. Smaller buffers mean lower latency but more callback overhead. The actual latencies in frames are reported by device.OutputLatencySamples after InitPlayback succeeds.

The buffer-switch callback runs on the ASIO driver's real-time thread, so any IWaveProvider in your chain must produce samples within the buffer duration. Allocations and slow I/O on that thread cause glitches.

Full example

c#
using NAudio.Wave;

using var reader = new AudioFileReader("music.wav");
using var device = AsioDevice.Open(AsioDevice.GetDriverNames()[0]);

device.InitPlayback(new AsioPlaybackOptions
{
    Source = reader,
    OutputChannels = [0, 1]
});

var done = new ManualResetEventSlim();
device.Stopped += (_, _) => done.Set();
device.Start();
done.Wait();

File: Docs/AsioRecording.md

Recording with ASIO

NAudio 3's AsioDevice exposes a clean recording mode that delivers per-channel Span<float> to your event handler — no IntPtr arithmetic, no manual sample-format decoding, no interleaving math. This article covers recording only — see AsioPlayback, AsioDuplex, and AsioMigration for related modes and migration guidance.

You need a soundcard with an ASIO driver installed. ASIO4ALL is a free fallback for hardware that doesn't ship one.

Open the device

c#
foreach (var name in AsioDevice.GetDriverNames())
    Console.WriteLine(name);

using var device = AsioDevice.Open("Focusrite USB ASIO");

Pick input channels

AsioRecordingOptions.InputChannels is an int[] of physical channel indices. Entries can be non-contiguous — record from channels [0, 1, 4, 5] directly without recording the channels in between.

c#
device.InitRecording(new AsioRecordingOptions
{
    InputChannels = [0, 1, 4, 5],
    SampleRate    = device.CurrentSampleRate
});

To record every available input:

c#
InputChannels = device.Capabilities.AllInputChannels

device.Capabilities.NbInputChannels reports how many physical inputs the driver exposes. device.Capabilities.InputChannelInfos[i].name gives the driver's name for each (e.g. "Mic 1", "Line 3").

If you don't pass SampleRate, the device runs at whatever rate the driver is currently set to — which device.CurrentSampleRate reports. If you pass a specific rate, the driver must support it; check first with device.IsSampleRateSupported(rate).

See AsioChannelMapping for the rationale and patterns around non-contiguous channel selection.

Subscribe to AudioCaptured

Each ASIO buffer-switch raises AudioCaptured on the real-time driver thread. The event args expose one ReadOnlySpan<float> per selected input, in the same order as InputChannels. NAudio handles the native AsioSampleType → float conversion (Int16LSB, Int24LSB, Int32LSB, Float32LSB are all supported transparently).

c#
device.AudioCaptured += (sender, e) =>
{
    // e.GetChannel(i) returns a ReadOnlySpan<float> for the i'th selected input.
    var ch0 = e.GetChannel(0);   // physical input InputChannels[0]
    var ch1 = e.GetChannel(1);   // physical input InputChannels[1]

    // Compute RMS, write to a file, push to a ring buffer — anything quick.
    // The spans are valid only for the duration of this handler.
};

device.Start();

The index passed to GetChannel is into the selected-channels array, not the physical channel number. If InputChannels = [4, 5], then GetChannel(0) returns physical input 4 and GetChannel(1) returns physical input 5. This matches JUCE/PortAudio conventions and keeps your handler portable across channel selections.

Real-time thread constraints

The handler runs on the ASIO callback thread. To avoid glitches:

  • Don't allocate. Pre-allocate any buffers you need before calling Start.
  • Don't perform blocking I/O. Hand the data off to a worker thread or a lock-free queue.
  • Don't call device.Stop, device.Dispose, or device.Reinitialize from inside the handler — Stop actively throws InvalidOperationException if you try, because the same-thread call would self-deadlock waiting for the callback to return.

The spans returned by GetChannel point into library-owned buffers that the device reuses across callbacks. Copy out anything you need to keep beyond the handler's return.

Save each input to its own WAV file

c#
using var device = AsioDevice.Open(driverName);
int[] channels = [0, 1, 4, 5];
int sampleRate = device.CurrentSampleRate;

device.InitRecording(new AsioRecordingOptions
{
    InputChannels = channels,
    SampleRate    = sampleRate
});

var writers = channels
    .Select(phys => new WaveFileWriter(
        $"input-{phys}.wav",
        WaveFormat.CreateIeeeFloatWaveFormat(sampleRate, 1)))
    .ToArray();

device.AudioCaptured += (s, e) =>
{
    for (int i = 0; i < e.ChannelCount; i++)
        writers[i].WriteSamples(e.GetChannel(i));
};

device.Stopped += (s, e) =>
{
    foreach (var w in writers) w.Dispose();
};

device.Start();
Console.ReadLine();
device.Stop();

Save selected inputs as a single multi-channel WAV

If you want to interleave the selected inputs into one WAV file:

c#
using var writer = new WaveFileWriter(
    "multi.wav",
    WaveFormat.CreateIeeeFloatWaveFormat(sampleRate, channels.Length));

device.AudioCaptured += (s, e) =>
{
    for (int frame = 0; frame < e.Frames; frame++)
        for (int ch = 0; ch < e.ChannelCount; ch++)
            writer.WriteSample(e.GetChannel(ch)[frame]);
};

The raw escape hatch

If you genuinely need zero-copy access to the driver's native bytes — for example, to memcpy them into a fixed buffer for a downstream codec — AsioAudioCapturedEventArgs.RawInput(i) returns an AsioRawInputBuffer:

c#
device.AudioCaptured += (s, e) =>
{
    var raw = e.RawInput(0);            // ref struct
    ReadOnlySpan bytes = raw.Bytes;
    AsioSampleType format = raw.Format; // Int16LSB / Int24LSB / Int32LSB / Float32LSB
    int frames = raw.Frames;
    // ... your zero-copy logic ...
};

The bytes span is valid only for the duration of the handler.

Stop and dispose

c#
device.Stop();
device.Dispose();   // or wrap the device in `using`

The Stopped event fires on the captured SynchronizationContext after the callback thread has fully drained, so it's safe for handlers to dispose the device.

File: Docs/ConcatenatingAudio.md

Concatenating Audio

When you play audio or render audio to a file, you create a single ISampleProvider or IWaveProvider that represents the whole piece of audio to be played. So playback will continue until you reach the end, and then stop.

But what if you have two pieces of audio you want to play back to back? The ConcatenatingSampleProvider enables you to schedule one or more pieces of audio to play one after the other.

Here's a simple example where we have three audio files that are going to play back to back. Note that the three audio files must have exactly the same sample rate, channel count and bit depth, because it's not possible to change those during playback.

c#
var first = new AudioFileReader("first.mp3");
var second = new AudioFileReader("second.mp3");
var third = new AudioFileReader("third.mp3");

var playlist = new ConcatenatingSampleProvider(new[] { first, second, third });

// to play:
outputDevice.Init(playlist);
outputDevice.Play();

// ... OR ... to save to file
WaveFileWriter.CreateWaveFile16("playlist.wav", playlist);

Note that the ConcatenatingSampleProvider does not provide repositioning. If you want that, you can quite simply copy the code for ConcatenatingSampleProvider and adjust it to allow you to rewind, or jump to the beginning of one of the inputs, depending on your specific requirements.

FollowedBy Extension Helpers

There are some helpful extension methods you can make use of to simplify concatenating. For example, to append one ISampleProvider onto the end of another, use FollowedBy. Under the hood this simply creates a ConcatenatingSampleProvider:

c#
var first = new AudioFileReader("first.mp3");
var second = new AudioFileReader("second.mp3");
var playlist = first.FollowedBy(second);

You can also provide a duration of silence that you want after the first sound has finished and before the second begins:

c#
var first = new AudioFileReader("first.mp3");
var second = new AudioFileReader("second.mp3");
var playlist = first.FollowedBy(TimeSpan.FromSeconds(1), second);

This makes use of an OffsetSampleProvider in conjunction with a ConcatenatingSampleProvider

File: Docs/ConvertBetweenStereoAndMono.md

Convert Between Stereo and Mono

NAudio includes a number of utility classes that can help you to convert between mono and stereo audio. You can use these whether you are playing audio live, or whether you are simply converting from one file format to another.

Mono to Stereo

If you have a mono input file, and want to convert to stereo, the MonoToStereoSampleProvider allows you to do this. It takes a SampleProvider as input, and has two floating point LeftVolume and RightVolume properties, which default to 1.0f. This means that the mono input will be copied at 100% volume into both left and right channels.

If you wanted to route it just to the left channel, you could set LeftVolume to 1.0f and RightVolume to 0.0f. And if you wanted it more to the right than the left you might set LeftVolume to 0.25f and RightVolume to 1.0f.

c#
using(var inputReader = new AudioFileReader(monoFilePath))
{
    // convert our mono ISampleProvider to stereo
    var stereo = new MonoToStereoSampleProvider(inputReader);
    stereo.LeftVolume = 0.0f; // silence in left channel
    stereo.RightVolume = 1.0f; // full volume in right channel

    // can either use this for playback:
    myOutputDevice.Init(stereo);
    myOutputDevice.Play();
    // ...

    // ... OR ... could write the stereo audio out to a WAV file
    WaveFileWriter.CreateWaveFile16(outputFilePath, stereo);
}

There's also a MonoToStereoProvider16 that works with 16 bit PCM IWaveProvider inputs and outputs 16 bit PCM. It works very similarly to MonoToStereoSampleProvider otherwise.

Stereo to Mono

If you have a stereo input file and want to collapse to mono, then the StereoToMonoSampleProvider is what you want. It takes a stereo ISampleProvider as input, and also has a LeftVolume and RightVolume property, although the defaults are 0.5f for each. This means the left sample will be multiplied by 0.5f and the right by 0.5f and the two are then summed together.

If you want to just keep the left channel and throw away the right, you'd set LeftVolume to 1.0f and RightVolume to 0.0f. You could even sort out an out of phase issue by setting LeftVolume to 0.5f and RightVolume to -0.5f.

Usage is almost exactly the same. Note that some output devices won't let you play a mono file directly, so this would be more common if you were creating a mono output file, or if the mono audio was going to be passed on as a mixer input to MixingSampleProvider.

c#
using(var inputReader = new AudioFileReader(stereoFilePath))
{
    // convert our stereo ISampleProvider to mono
    var mono = new StereoToMonoSampleProvider(inputReader);
    mono.LeftVolume = 0.0f; // discard the left channel
    mono.RightVolume = 1.0f; // keep the right channel

    // can either use this for playback:
    myOutputDevice.Init(mono);
    myOutputDevice.Play();
    // ...

    // ... OR ... could write the mono audio out to a WAV file
    WaveFileWriter.CreateWaveFile16(outputFilePath, mono);
}

There is also a StereoToMonoProvider16 that works with 16 bit PCM stereo IWaveProvider inputs and emits 16 bit PCM.

Panning Mono to Stereo

Finally, NAudio offers a PanningSampleProvider which allows you to use customisable panning laws to govern how a mono input signal is placed into a stereo output signal.

It has a Pan property which can be configured between -1.0f (fully left) and 1.0f (fully right), with 0.0f being central.

The PanningStrategy can be overridden. By default is uses the SinPanStrategy. There is also SquareRootPanStrategy, LinearPanStrategy and StereoBalanceStrategy, each one operating slightly differently with regards to how loud central panning is, and how the sound tapers off as it is panned to each side. You can experiment to discover which one fits your needs the best.

Usage is very similar to the MonoToStereoSampleProvider

c#
using(var inputReader = new AudioFileReader(monoFilePath))
{
    // convert our mono ISampleProvider to stereo
    var panner = new PanningSampleProvider(inputReader);
    // override the default pan strategy
    panner.PanStrategy = new SquareRootPanStrategy();
    panner.Pan = -0.5f; // pan 50% left

    // can either use this for playback:
    myOutputDevice.Init(panner);
    myOutputDevice.Play();
    // ...

    // ... OR ... could write the stereo audio out to a WAV file
    WaveFileWriter.CreateWaveFile16(outputFilePath, panner);
}

File: Docs/ConvertMp3ToWav.md

Convert an MP3 File to a WAV File

In this article I will show a few ways you can convert an MP3 file into a WAV file with NAudio.

To start with we'll need a couple of file paths, one to the input MP3 file, and one to where we want to put the converted WAV file.

c#
var infile = @"C:\Users\Mark\Desktop\example.mp3";
var outfile = @"C:\Users\Mark\Desktop\converted.wav";

MediaFoundationReader is the recommended approach for reading MP3 files (and many other formats) in NAudio. It uses Media Foundation which is available on all supported versions of Windows. It can read MP3, WMA, AAC, FLAC, Opus and many other formats.

c#
using(var reader = new MediaFoundationReader(infile))
{
    WaveFileWriter.CreateWaveFile(outfile, reader);
}

Mp3FileReader

The Mp3FileReader class uses the ACM MP3 codec that is present on most versions of Windows. The conversion is straightforward. Open the MP3 file with Mp3FileReader and then pass it to WaveFileWriter.CreateWaveFile to write the converted PCM audio to a WAV file. This will usually be 44.1kHz 16 bit stereo, but uses whatever format the MP3 decoder emits.

c#
using(var reader = new Mp3FileReader(infile))
{
    WaveFileWriter.CreateWaveFile(outfile, reader);
}

DirectX Media Object

Mp3FileReaderBase allows us to plug in alternative MP3 frame decoders. One option that comes in the box with NAudio is the DirectX Media Object MP3 codec.

Here's how to use the DmoMp3FrameDecompressor as a custom frame decompressor:

c#
using(var reader = new Mp3FileReaderBase(infile, wf => new DmoMp3FrameDecompressor(wf)))
{
    WaveFileWriter.CreateWaveFile(outfile, reader);
}

NLayer

The final option is to use NLayer as the decoder for Mp3FileReader. NLayer is a fully managed MP3 decoder, meaning it can run on any .NET platform including cross-platform scenarios where Windows codecs are not available. You'll need the NLayer.NAudioSupport NuGet package. Then you can plug in a fully managed MP3 frame decoder:

c#
using (var reader = new Mp3FileReaderBase(infile, wf => new Mp3FrameDecompressor(wf)))
{
    WaveFileWriter.CreateWaveFile(outfile, reader);
}

File: Docs/EnumerateAcmDrivers.md

Enumerate ACM Drivers

ACM drivers are the old Windows API for dealing with compressed audio that predates Media Foundation. In one sense this means that this is no longer very important, but sometimes you find that some codecs are more readily available as ACM codecs instead of Media Foundation Transforms.

The class in NAudio that makes use of ACM codecs is WaveFormatConversionStream. When you construct one you provide it with a source and a target WaveFormat. This will be either going from compressed audio to PCM (this is a decoder) or from PCM to compressed (this is an encoder). Its important to not that you can't just pick two random WaveFormat definitions and expect a conversion to be possible. You can only perform the supported transforms.

That's why it's really useful to be able to enumerate the ACM codecs installed on your system. You can do that with AcmDriver.EnumerateAcmDrivers. Then you explore the FormatTags for each driver, and from there ask for each format matching that tag with driver.GetFormats.

It is a little complex, but the information you get from doing this is invaluable in helping you to work out exactly what WaveFormat you need to use to successfully use a codec.

This code sample enumerates through all ACM drivers and prints out details of their formats.

text
/* Detailed source-code truncated for AI context efficiency. */

The output will be quite verbose (especially if you've installed some additional codecs on your system.) Here's a snippet of the output from the GSM codec:

text
Long Name: Microsoft GSM 6.10 Audio CODEC
Short Name: Microsoft GSM 6.10
Driver ID: 48141232
FormatTags:
===========================================
Format Tag 0: PCM
   Standard Format Count: 8
   Support Flags: Codec
   Format Tag: Pcm, Format Size: 16
   Formats:
   ===========================================
   Format 0: 8.000 kHz, 8 Bit, Mono
      FormatTag: Pcm, Support Flags: Codec
      WaveFormat: Pcm 8000Hz Channels: 1 Bits: 8 Block Align: 1, AverageBytesPerSecond: 8000 (64.0 kbps), Extra Size: 0
   ===========================================
   Format 1: 8.000 kHz, 16 Bit, Mono
      FormatTag: Pcm, Support Flags: Codec
      WaveFormat: Pcm 8000Hz Channels: 1 Bits: 16 Block Align: 2, AverageBytesPerSecond: 16000 (128.0 kbps), Extra Size: 0
   ===========================================
   Format 2: 11.025 kHz, 8 Bit, Mono
      FormatTag: Pcm, Support Flags: Codec
      WaveFormat: Pcm 11025Hz Channels: 1 Bits: 8 Block Align: 1, AverageBytesPerSecond: 11025 (88.2 kbps), Extra Size: 0

And here's an example showing a non-PCM format. Here we can see that for DviAdpcm, the WaveFormat structure needs two extra bytes with values 0xF9 and 0x01:

text
===========================================
   Format 1: 8.000 kHz, 4 Bit, Stereo
      FormatTag: DviAdpcm, Support Flags: Codec
      WaveFormat: DviAdpcm 8000Hz Channels: 2 Bits: 4 Block Align: 512, AverageBytesPerSecond: 8110 (64.9 kbps), Extra Size: 2
      Extra Bytes:
      F9 01

File: Docs/EnumerateMediaFoundationTransforms.md

Enumerate Media Foundation Transforms

The MediaFoundationReader and MediaFoundationEncoder classes in NAudio make use of any available Media Foundation Transforms (MFTs) installed on your computer. It can be useful to enumerate the available audio-related MFTs.

There are three categories of audio MFT - effects, decoders and encoders. A decoder allows you to decode audio compressed in different formats to PCM. An encoder allows you to encode PCM audio into compressed formats. An effect modifies audio in some way (e.g. resampling).

You can use MediaFoundationApi.EnumerateTransforms to explore what's available:

c#
MediaFoundationApi.Startup();

var effects = MediaFoundationApi.EnumerateTransforms(MediaFoundationTransformCategories.AudioEffect);
var decoders = MediaFoundationApi.EnumerateTransforms(MediaFoundationTransformCategories.AudioDecoder);
var encoders = MediaFoundationApi.EnumerateTransforms(MediaFoundationTransformCategories.AudioEncoder);

These return an IEnumerable<MfActivate>. The MfActivate wrapper provides methods to explore the attributes of each transform:

text
/* Detailed source-code truncated for AI context efficiency. */

Here's an example output for an MFT effect. In this case, the Resampler which is a very useful MFT for changing sample rates:

text
Audio Effect
Name: Resampler MFT
Input Types: 2 items:
    Audio-PCM
    Audio-IEEE floating-point
Class identifier: f447b69e-1884-4a7e-8055-346f74d6edb3
Output Types: 2 items:
    Audio-PCM
    Audio-IEEE floating-point
Transform Flags: 1
Transform Category: Audio Effect

Here's an example output for a decoder. Windows 10 and above includes an Opus audio decoder:

text
Audio Decoder
Name: Microsoft Opus Audio Decoder MFT
Input Types: 1 items:
    Audio-Opus
Class identifier: 63e17c10-2d43-4c42-8fe3-8d8b63e46a6a
Output Types: 1 items:
    Audio-IEEE floating-point
Transform Flags: 1
Transform Category: Audio Decoder

And an encoder. Windows 10 and above includes a FLAC encoder:

text
Audio Encoder
Name: Microsoft FLAC Audio Encoder MFT
Input Types: 1 items:
    Audio-PCM
Class identifier: 128509e9-c44e-45dc-95e9-c255b8f466a6
Output Types: 1 items:
    Audio-FLAC
Transform Flags: 1
Transform Category: Audio Encoder

You can also activate a transform directly from the MfActivate object. For example, to create an MfTransform wrapper:

c#
using var transform = mft.ActivateTransform();

File: Docs/EnumerateOutputDevices.md

Enumerating Audio Devices

The technique you use to enumerate audio devices depends on what audio output (or input) driver type you are using. This article shows the technique for each supported output device.

WaveOut

To discover the number of output devices you can use WaveOut.DeviceCount. Then you can call WaveOut.GetCapabilities passing in the index of a device to find out its name (and some basic information about its capabilities).

Note that you can also pass an index of -1 which is the "audio mapper". Use this if you want to keep playing audio even when a device is removed (such as USB headphones being unplugged).

Also note that the ProductName retured is limited to 32 characters, resulting in it often being truncated. This is a limitation of the underlying Windows API and there is unfortunately no easy way to fix it in NAudio.

c#
for (int n = -1; n < WaveOut.DeviceCount; n++)
{
    var caps = WaveOut.GetCapabilities(n);
    Console.WriteLine($"{n}: {caps.ProductName}");
}

Once you've selected the device you want, you can open it by creating an instance of WaveOut and specifying it as the DeviceNumber:

c#
var outputDevice = new WaveOut() { DeviceNumber = deviceNumber };

WaveIn

Getting details of audio capture devices for WaveIn is very similar to for WaveOut:

c#
for (int n = -1; n < WaveIn.DeviceCount; n++)
{
    var caps = WaveIn.GetCapabilities(n);
    Console.WriteLine($"{n}: {caps.ProductName}");
}

Once you've selected the device you want, you can open it by creating an instance of WaveIn and specifying it as the DeviceNumber:

c#
var recordingDevice = new WaveIn() { DeviceNumber = deviceNumber };

DirectSoundOut

DirectSoundOut exposes the Devices static method allowing you to enumerate through all the output devices. This has the benefit over WaveOut of not having truncated device names:

c#
foreach (var dev in DirectSoundOut.Devices)
{
    Console.WriteLine($"{dev.Guid} {dev.ModuleName} {dev.Description}");
}

Each device has a Guid, and that can be used to open a specific device:

c#
var outputDevice = new DirectSoundOut(deviceGuid);

There are also a couple of special device GUIDs you can use to open the default playback device (DirectSoundOut.DSDEVID_DefaultPlayback) or default voice playback device (DirectSoundOut.DSDEVID_DefaultVoicePlayback)

WASAPI Devices

WASAPI playback (render) and recording (capture) devices can both be accessed via the MMDeviceEnumerator class. This allows you to enumerate only the type of devices you want (DataFlow.Render or DataFlow.Capture or DataFlow.All).

You can also choose whether you want to include devices that are active, or also include disabled, unplugged or otherwise not present devices with the DeviceState bitmask. Here we show them all:

c#
var enumerator = new MMDeviceEnumerator();
foreach (var wasapi in enumerator.EnumerateAudioEndPoints(DataFlow.All, DeviceState.All))
{
    Console.WriteLine($"{wasapi.DataFlow} {wasapi.FriendlyName} {wasapi.DeviceFriendlyName} {wasapi.State}");
}

To open the device you want, pass the device in to the appropriate WASAPI builder depending on whether you are playing back or recording...

c#
var outputDevice = new WasapiPlayerBuilder().WithDevice(mmDevice).Build();
var recordingDevice = new WasapiRecorderBuilder().WithDevice(captureDevice).Build();
var loopbackCapture = new WasapiRecorderBuilder().WithDevice(loopbackDevice).WithLoopbackCapture().Build();

You can also use the MMEnumerator to request what the default device is for a number of different scenarios (playback or record, and voice, multimedia or 'console'):

c#
enumerator.GetDefaultAudioEndpoint(DataFlow.Render, Role.Multimedia);

ASIO

You can discover the registered ASIO drivers on your system with AsioDevice.GetDriverNames. There is no guarantee that the associated soundcard is currently connected to the system.

c#
foreach (var asio in AsioDevice.GetDriverNames())
{
    Console.WriteLine(asio);
}

You can then use the driver name to open the device:

c#
using var device = AsioDevice.Open(driverName);

(AsioDevice is the NAudio 3 ASIO API. The legacy AsioOut class still works — see Migrating from AsioOut to AsioDevice.)

Management Objects

Finally you can use Windows Management Objects to get hold of details of the sound devices installed. This doesn't map specifically to any of the NAudio output device types, but can be a source of useful information

c#
var objSearcher = new ManagementObjectSearcher(
	   "SELECT * FROM Win32_SoundDevice");

var objCollection = objSearcher.Get();
foreach (var d in objCollection)
{
    Console.WriteLine("=====DEVICE====");
    foreach (var p in d.Properties)
    {
        Console.WriteLine($"{p.Name}:{p.Value}");
    }
}

File: Docs/FadeInOutSampleProvider.md

Fading Audio in and out with FadeInOutSampleProvider

The FadeInOutSampleProvider offers a simple, basic way to fade audio in and out.

It follows the decorator pattern common to many ISampleProvider implementations. You pass in the ISampleProvider that you want to fade in and out.

In this example, we'll construct a FadeInOutSampleProvider taking its source from an AudioFileReader, and passing the true flag to specify that we want to start with silence, ready for a fade in.

We'll also immediately trigger a fade in over 2 seconds (2000 milliseconds) by calling BeginFadeIn.

c#
var audio = new AudioFileReader("example.mp3");
var fade = new FadeInOutSampleProvider(audio, true);
fade.BeginFadeIn(2000);

Now we can pass our FadeInOutSampleProvider to an output device and start playing. We'll hear the audio fading in over the first two seconds.

c#
var waveOutDevice = new WaveOut();
waveOutDevice.Init(fade);
waveOutDevice.Play();

At some point in the future, we might want to fade out, and we can trigger that with BeginFadeOut, again specifying a 2 second fadeout.

c#
fade.BeginFadeOut(2000);

Once the audio has faded out, the FadeInOutSampleProvider continues to read from its source but emits silence until it reaches its end, or until you call BeginFadeIn again.

Taking it further

The FadeInOutSampleProvider is a very basic fade provider, and you may want additional features like:

  • automatically fading out when you reach the end of the source
  • automatically stopping at the end of a fade out
  • cross-fading into another input.

You can do this by taking the code for FadeInOutSampleProvider and adapting it.

For example, to automatically fade out at the end of the source, you'd actually need to read ahead by the duration of the fade (or know in advance where you want the fade to begin)

These features may be added in the future to NAudio, but don't be afraid to create your own custom ISampleProvider implementations that behave just how you want.

Explore Related Technologies & Topics

View All 9,111+ Repositories →

Discover similar high-velocity repositories, agent skills, and OpenAPI specifications across the ecosystem.

⚡ MCP Servers Model Context Protocol → 🧠 Agent Skills Prompt & rule files → 🔌 OpenAPI Specs REST & Swagger schemas →