通过 Yeti lame 包装器将音频流式传输到 mp3 文件
Posted
技术标签:
【中文标题】通过 Yeti lame 包装器将音频流式传输到 mp3 文件【英文标题】：NAudio streaming to mp3 file throught yeti lame wrapper 【发布时间】：2013-09-22 20:52:30 【问题描述】：我有以下代码来记录音频输入和输出
using System;
using System.Diagnostics;
using System.IO;
using NAudio.Wave;
using Yeti.MMedia.Mp3;
namespace SoundRecording
public class SoundManager
private WaveInEvent _waveIn;
private WaveFileWriter _waveInFile;
private WasapiLoopbackCapture _waveOut;
private WaveFileWriter _waveOutFile;
private Process _lameProcess;
public void StartRecording()
InitLame();
DateTime dtNow = DateTime.Now;
try
InitAudioOut(dtNow);
catch
try
InitAudioIn(dtNow);
catch
private void InitLame()
string outputFileName = @"c:\Rec\test.mp3";
_lameProcess = new Process();
_lameProcess.StartInfo.FileName = @"lame.exe";
_lameProcess.StartInfo.UseShellExecute = false;
_lameProcess.StartInfo.RedirectStandardInput = true;
_lameProcess.StartInfo.Arguments = "-r -s 44.1 -h -b 256 --bitwidth 32 - \"" + outputFileName + "\"";
_lameProcess.StartInfo.CreateNoWindow = true;
_lameProcess.Start();
private void InitAudioIn(DateTime dtNow)
string pathIn = @"C:\Rec\(" + dtNow.ToString("HH-mm-ss") + " " + dtNow.ToString("dd-MM-yyyy") + " IN).wav";
_waveIn = new WaveInEvent();
_waveIn.WaveFormat = new WaveFormat(8000, 1);
_waveIn.DataAvailable += WaveInDataAvailable;
_waveIn.RecordingStopped += WaveInRecordStopped;
_waveInFile = new WaveFileWriter(pathIn, _waveIn.WaveFormat);
_waveIn.StartRecording();
private void InitAudioOut(DateTime recordMarker)
string pathOut = @"C:\Rec\(" + recordMarker.ToString("HH-mm-ss") + " " + recordMarker.ToString("dd-MM-yyyy") + " OUT).mp3";
_waveOut = new WasapiLoopbackCapture();
//_waveOut.WaveFormat = new WaveFormat(44100, 1);
_waveOut.DataAvailable += WaveOutDataAvailable;
_waveOut.RecordingStopped += WaveOutRecordStopped;
_waveOutFile = new WaveFileWriter(pathOut, new Mp3WaveFormat(_waveOut.WaveFormat.SampleRate, _waveOut.WaveFormat.Channels, 0, 128));
_waveOut.StartRecording();
private void WaveInDataAvailable(object sender, WaveInEventArgs e)
if (_waveInFile != null)
_waveInFile.Write(e.Buffer, 0, e.BytesRecorded);
_waveInFile.Flush();
private void WaveOutDataAvailable(object sender, WaveInEventArgs e)
if (_waveInFile != null)
using (var memStream = new MemoryStream(e.Buffer))
using (WaveStream wStream = new RawSourceWaveStream(memStream, _waveOut.WaveFormat))
var format = new WaveFormat(_waveOut.WaveFormat.SampleRate, _waveOut.WaveFormat.Channels);
var transcodedStream = new ResamplerDmoStream(wStream, format);
var read = (int)transcodedStream.Length;
var bytes = new byte[read];
transcodedStream.Read(bytes, 0, read);
var fmt = new WaveLib.WaveFormat(transcodedStream.WaveFormat.SampleRate, transcodedStream.WaveFormat.BitsPerSample, transcodedStream.WaveFormat.Channels);
var beconf = new Yeti.Lame.BE_CONFIG(fmt, 128);
// Encode WAV to MP3
byte[] mp3Data;
using (var mp3Stream = new MemoryStream())
using (var mp3Writer = new Mp3Writer(mp3Stream, fmt, beconf))
int blen = transcodedStream.WaveFormat.AverageBytesPerSecond;
mp3Writer.Write(bytes, 0, read);
mp3Data = mp3Stream.ToArray();
_waveOutFile.Write(mp3Data, 0, mp3Data.Length);
_waveOutFile.Flush();
private byte[] WavBytesToMp3Bytes(IWaveProvider waveStream, uint bitrate = 128)
// Setup encoder configuration
var fmt = new WaveLib.WaveFormat(waveStream.WaveFormat.SampleRate, waveStream.WaveFormat.BitsPerSample, waveStream.WaveFormat.Channels);
var beconf = new Yeti.Lame.BE_CONFIG(fmt, bitrate);
// Encode WAV to MP3
int blen = waveStream.WaveFormat.AverageBytesPerSecond;
var buffer = new byte[blen];
byte[] mp3Data = null;
using (var mp3Stream = new MemoryStream())
using (var mp3Writer = new Mp3Writer(mp3Stream, fmt, beconf))
int readCount;
while ((readCount = waveStream.Read(buffer, 0, blen)) > 0)
mp3Writer.Write(buffer, 0, readCount);
mp3Data = mp3Stream.ToArray();
return mp3Data;
private void WaveInRecordStopped(object sender, StoppedEventArgs e)
if (_waveIn != null)
_waveIn.Dispose();
_waveIn = null;
if (_waveInFile != null)
_waveInFile.Dispose();
_waveInFile = null;
_lameProcess.StandardInput.BaseStream.Close();
_lameProcess.StandardInput.BaseStream.Dispose();
_lameProcess.Close();
_lameProcess.Dispose();
private void WaveOutRecordStopped(object sender, StoppedEventArgs e)
if (_waveOutFile != null)
_waveOutFile.Close();
_waveOutFile = null;
_waveOut = null;
public void StopRecording()
try
_waveIn.StopRecording();
catch
try
_waveOut.StopRecording();
catch
我正在使用 NAudio 捕获音频输入/输出，并且使用 Yetis 的蹩脚包装器将其即时转换为 mp3 文件，问题是生成的音频输出文件已损坏且无法读取，可能缺少 mp3 标头或其他东西其他我错过的......
【问题讨论】：
【参考方案1】：问题是您从环回捕获接口以默认格式（即：PCM）获取批量数据，然后将其写入带有 声称的格式块的波形文件数据为 ALAW 格式。在任何时候，您实际上都没有进行从 PCM 数据到 ALAW 数据的转换，从而产生了一个垃圾文件。
WaveFileWriter 类不会为您进行任何形式的重新编码或重新采样。它使用格式说明符为 WAV 文件构建格式块，并假定您正在向它提供该格式的数据。
您的两个选择是：
在写入WaveFileWriter 实例之前，将传入的数据从 PCM-44100-Stereo（或任何默认值）转换为 ALAW-8000-Mono。
用_waveOut.WaveFormat初始化_waveOutFile以匹配数据格式。
9 月 26 日更新...
因此，经过一番折腾，我终于有了一个可行的解决方案来解决原来的问题，即正确地将来自环回捕获的波形格式转换为可以压缩的格式。
这是转换第一阶段的代码：
[StructLayout(LayoutKind.Explicit)]
internal struct UnionStruct
[FieldOffset(0)]
public byte[] bytes;
[FieldOffset(0)]
public float[] floats;
public static byte[] Float32toInt16(byte[] data, int offset, int length)
UnionStruct u = new UnionStruct();
int nSamples = length / 4;
if (offset == 0)
u.bytes = data;
else
u.bytes = new byte[nSamples * 4];
Buffer.BlockCopy(data, offset, u.bytes, 0, nSamples * 4);
byte[] res = new byte[nSamples * 2];
for (i = 0, o = 0; i < nSamples; i++, o+= 2)
short val = (short)(u.floats[i] * short.MaxValue);
res[o] = (byte)(val & 0xFF);
res[o + 1] = (byte)((val >> 8) & 0xFF);
u.bytes = null;
return res;
这会将 32 位浮点样本转换为可由大多数音频代码处理的 16 位有符号整数样本。幸运的是，这包括Yeti MP3 代码。
要即时编码并确保 MP3 输出有效，请同时创建 Mp3Writer 及其输出 Stream（例如，FileStream 直接写入磁盘）并保持从环回接口输入数据（通过上面的转换器运行）。在waveInStopRecording 事件处理程序中关闭Mp3Writer 和Stream。
Stream _mp3Output;
Mp3Writer _mp3Writer;
private void InitAudioOut(DateTime recordMarker)
string pathOut = string.Format(@"C:\Rec\(0:HH-mm-ss dd-MM-yyyy OUT).mp3", recordMarker);
_waveOut = new WasapiLoopbackCapture();
_waveOut.DataAvailable += WaveOutDataAvailable;
_waveOut.RecordingStopped += WaveOutRecordStopped;
_mp3Output = File.Create(pathIn);
var fmt = new WaveLib.WaveFormat(_waveOut.WaveFormat.SampleRate, 16, _waveOut.Channels);
var beconf = new Yeti.Lame.BE_CONFIG(fmt, 128);
_mp3Writer = new Mp3Writer(_mp3Stream, fmt, beconf);
_waveOut.StartRecording();
private void WaveOutDataAvailable(object sender, WaveInEventArgs e)
if (_mp3Writer != null)
byte[] data = Float32toInt16(e.Buffer, 0, e.BytesRecorded);
_mp3Writer.Write(data, 0, data.Length);
private void WaveOutRecordStopped(object sender, StoppedEventArgs e)
if (InvokeRequired)
BeginInvoke(new MethodInvoker(WaveOutStop));
else
WaveOutStop();
private void WaveOutStop()
if (_mp3Writer != null)
_mp3Writer.Close();
_mp3Writer.Dispose();
_mp3Writer = null;
if (_mp3Stream != null)
_mp3Stream.Dispose();
_mp3Stream = null;
_waveOut.Dispose();
_waveOut = null;
顺便说一句，Mp3Writer 类就是你所需要的。扔掉你那里的其他Lame 代码。它只会妨碍你。
【讨论】：
第二个工作正常，但它会产生一个巨大的文件（这就是我一开始拒绝它的原因）。您能否提供第一个选项的代码示例？我正在尝试这样做，因为它写在帖子中，但似乎我在某个地方错了或错过了什么，提前谢谢你 @HardLuck 应该很简单，但目前压缩对我来说有点搞砸了。我一直在使用 Yeti 的蹩脚包装器来创建 MP3 文件。 遗憾的是我不知道如何使用它，但似乎我应该尝试一下，谢谢您 嘿，那里有很棒的更新！但是你能告诉我 Float32toInt16 方法中的“v”变量是什么吗？ @HardLuck 是的，这是一个错字：P 当我在这里输入代码而不是从我的工作代码中复制粘贴时，就会发生这种情况。现在修好了。【参考方案2】：WasapiLoopbackCapture 可能会以 32 位浮点、44.1kHz、立体声捕获音频。 WaveFormatConversionStream 不会一步将其转换为 8kHz 单声道。您需要分多个步骤进行此转换。
【讨论】：
感谢您的回答，这让情况更加清晰，我已经更新了我的问题以匹配其当前状态以上是关于通过 Yeti lame 包装器将音频流式传输到 mp3 文件的主要内容，如果未能解决你的问题，请参考以下文章