Update your bookmarks! This blog is now hosted on http://xoofx.com/blog

Saturday, May 22, 2010

Playing a MP3 in c++ using plain Windows API

While playing a mp3 is quite common in a demo, I have seen that most demo are often using 3rd party dlls like Bass or FMod to perform this simple task under windows. But if we want to get rid off this dependency, how can we achieve this with a plain windows API? What's the requirements to have a practical MP3Player for a demo?

Surprisingly, I was not able to find a simple code sample other the Internet that explain how to play a mp3 with Windows API, without using the too simple Windows Media Player API. Why WMP is not enough (not even talking about MCI - Media Control Interface which is even more basic than WMP)?

Well, It's lacking one important feature : It's only able to play from an url, so it's not possible for example to pack the song in an archive and play it from a memory location (although not a huge deal if you want to release the song on the side of your demo). Also I have never tested the timing returned by WMP (using probably IWMPControls3 getPositionTimeCode) and not really sure It's able to provide a reliable sync (at least, If you intend to use sync... but hey, is a demo without any sync, can be still called a demo?:)

So I started to find some peace of code around the net but they were covering only part of the problem. The starting point was to rely on the Audio Compression Manager API that provides an API conversion to perform for example a mp3 to pcm. Hopefully, I found the code from a guy that was kind enough to post the whole converter for a mp3 file using ACM. In the mean time, I found that Mark Heath, the author of NAudio posted few days ago a solution to convert a MP3 to WAV using NAudio. Looking at his code, he was using also ACM but he reported also some difficulties to implement a reliable MP3Frame/ID3Tag decoder in order to extract samplerate, bitrate, channels...etc. I didn't want to use this kind of heavy code and was looking a lighter and reliable solution for this : most of the people were talking about using the Windows Media Format SDK to get all this information from the file. The starting point is the WMCreateSyncReader method. Through this method, you are able to retrieve part of MP3Frame as well as ID3Tag.

Finally, I came with a patchwork solution :
  • using SyncReader from WMF to extract song duration.
  • using ACM to decode the mp3 to pcm
  • using plain old waveOut functions to perform sound playback and retrieve sound playback position.
Everything is inside a single .h with less than 300 lines including comments. I don't really know If it's the best way to play a mp3 from a file or from the memory, with Windows API, while still providing a reliable timing. I have tested it against a couple of mp3, thus It may still have some bugs... but at least, It's working quite well and It's a pretty small code. For example, the code is expecting the input mp3 to be at 44100Hz samplerate... If not It should probably failed... although with the use of WMF, It's quite easy to extract the sampleRate (although I'm not using it in the sample code provided here... was not sure about the result though :) )

Also, the code is not decoding&playing in realtime the song but is instead performing the decoding in a single pass and then playing the decoded buffer. This requires that the full pcm song to be allocated, which could be around 20Mo to 50Mo depending on the size of your song (It's easy to calculate : durationInSecondsOfTheSong * 4 * 441000, so a 3min song is requiring 30Mo). This is not probably the best solution, but It's not a huge task to transform this code to do realtime decoding/playback. The downside is that It will take some CPU in your demo. So that in the end, It's a just tradeoff between memory vs cpu depending on your needs!

/* ----------------------------------------------------------------------
 * MP3Player.h C++ class using plain Windows API
 *
 * Author: @lx/Alexandre Mutel,  blog: http://code4k.blogspot.com
 * The software is provided "as is", without warranty of any kind.
 * ----------------------------------------------------------------------*/
#pragma once
#include <windows.h>
#include <stdio.h>
#include <assert.h>
#include <mmreg.h>
#include <msacm.h>
#include <wmsdk.h>

#pragma comment(lib, "msacm32.lib") 
#pragma comment(lib, "wmvcore.lib") 
#pragma comment(lib, "winmm.lib") 
#pragma intrinsic(memset,memcpy,memcmp)

#ifdef _DEBUG
#define mp3Assert(function) assert((function) == 0)
#else
//#define mp3Assert(function) if ( (function) != 0 ) { MessageBoxA(NULL,"Error in [ " #function "]", "Error",MB_OK); ExitProcess(0); }
#define mp3Assert(function) (function)
#endif

/*
 * MP3Player class.
 * Usage : 
 *   MP3Player player;
 *   player.OpenFromFile("your.mp3");
 *   player.Play();
 *   Sleep((DWORD)(player.GetDuration()+1));
 *   player.Close();
 */
class MP3Player {
private:
 HWAVEOUT hWaveOut;
 DWORD bufferLength;
 double durationInSecond;
 BYTE* soundBuffer;
public:

 /*
  * OpenFromFile : loads a MP3 file and convert it internaly to a PCM format, ready for sound playback.
  */
 HRESULT OpenFromFile(TCHAR* inputFileName){
  // Open the mp3 file
  HANDLE hFile = CreateFile(inputFileName, // open MYFILE.TXT
         GENERIC_READ,
         FILE_SHARE_READ, // share for reading
         NULL, // no security
         OPEN_EXISTING, // existing file only
         FILE_ATTRIBUTE_NORMAL, // normal file
         NULL); // no attr
  assert( hFile != INVALID_HANDLE_VALUE);

  // Get FileSize
  DWORD fileSize = GetFileSize(hFile, NULL);
  assert( fileSize != INVALID_FILE_SIZE);

  // Alloc buffer for file
  BYTE* mp3Buffer = (BYTE*)LocalAlloc(LPTR, fileSize);

  // Read file and fill mp3Buffer
  DWORD bytesRead;
  DWORD resultReadFile = ReadFile( hFile, mp3Buffer, fileSize, &bytesRead, NULL);
   assert(resultReadFile != 0);
  assert( bytesRead == fileSize);

  // Close File
  CloseHandle(hFile);

  // Open and convert MP3
  HRESULT hr = OpenFromMemory(mp3Buffer, fileSize);

  // Free mp3Buffer
  LocalFree(mp3Buffer);

  return hr;
 }

 /*
  * OpenFromMemory : loads a MP3 from memory and convert it internaly to a PCM format, ready for sound playback.
  */
 HRESULT OpenFromMemory(BYTE* mp3InputBuffer, DWORD mp3InputBufferSize){
  IWMSyncReader* wmSyncReader;
  IWMHeaderInfo* wmHeaderInfo;
  IWMProfile* wmProfile;
  IWMStreamConfig* wmStreamConfig;
  IWMMediaProps* wmMediaProperties;
  WORD wmStreamNum = 0;
  WMT_ATTR_DATATYPE wmAttrDataType;
  DWORD durationInSecondInt;
  QWORD durationInNano;
  DWORD sizeMediaType;
  DWORD maxFormatSize = 0;
  HACMSTREAM acmMp3stream = NULL;
  HGLOBAL mp3HGlobal;
  IStream* mp3Stream;

  // Define output format
  static WAVEFORMATEX pcmFormat = {
   WAVE_FORMAT_PCM, // WORD        wFormatTag;         /* format type */
   2,     // WORD        nChannels;          /* number of channels (i.e. mono, stereo...) */
   44100,    // DWORD       nSamplesPerSec;     /* sample rate */
   4 * 44100,   // DWORD       nAvgBytesPerSec;    /* for buffer estimation */
   4,     // WORD        nBlockAlign;        /* block size of data */
   16,     // WORD        wBitsPerSample;     /* number of bits per sample of mono data */
   0,     // WORD        cbSize;             /* the count in bytes of the size of */
  };

  const DWORD MP3_BLOCK_SIZE = 522;

  // Define input format
  static MPEGLAYER3WAVEFORMAT mp3Format = {
   {
    WAVE_FORMAT_MPEGLAYER3,   // WORD        wFormatTag;         /* format type */
     2,        // WORD        nChannels;          /* number of channels (i.e. mono, stereo...) */
     44100,       // DWORD       nSamplesPerSec;     /* sample rate */
     128 * (1024 / 8),    // DWORD       nAvgBytesPerSec;    not really used but must be one of 64, 96, 112, 128, 160kbps
     1,        // WORD        nBlockAlign;        /* block size of data */
     0,        // WORD        wBitsPerSample;     /* number of bits per sample of mono data */
     MPEGLAYER3_WFX_EXTRA_BYTES,  // WORD        cbSize;        
   },
   MPEGLAYER3_ID_MPEG,      // WORD          wID;
   MPEGLAYER3_FLAG_PADDING_OFF,   // DWORD         fdwFlags;
   MP3_BLOCK_SIZE,       // WORD          nBlockSize;
   1,          // WORD          nFramesPerBlock;
   1393,         // WORD          nCodecDelay;
  };

  // -----------------------------------------------------------------------------------
  // Extract and verify mp3 info : duration, type = mp3, sampleRate = 44100, channels = 2
  // -----------------------------------------------------------------------------------

  // Initialize COM
  CoInitialize(0);

  // Create SyncReader
  mp3Assert( WMCreateSyncReader(  NULL, WMT_RIGHT_PLAYBACK , &wmSyncReader ) );

  // Alloc With global and create IStream
  mp3HGlobal = GlobalAlloc(GPTR, mp3InputBufferSize);
  assert(mp3HGlobal != 0);
  void* mp3HGlobalBuffer = GlobalLock(mp3HGlobal);
  memcpy(mp3HGlobalBuffer, mp3InputBuffer, mp3InputBufferSize);
  GlobalUnlock(mp3HGlobal);
  mp3Assert( CreateStreamOnHGlobal(mp3HGlobal, FALSE, &mp3Stream) );

  // Open MP3 Stream
  mp3Assert( wmSyncReader->OpenStream(mp3Stream) );

  // Get HeaderInfo interface
  mp3Assert( wmSyncReader->QueryInterface(&wmHeaderInfo) );

  // Retrieve mp3 song duration in seconds
  WORD lengthDataType = sizeof(QWORD);
  mp3Assert( wmHeaderInfo->GetAttributeByName(&wmStreamNum, L"Duration", &wmAttrDataType, (BYTE*)&durationInNano, &lengthDataType ) );
  durationInSecond = ((double)durationInNano)/10000000.0;
  durationInSecondInt = (int)(durationInNano/10000000)+1;

  // Sequence of call to get the MediaType
  // WAVEFORMATEX for mp3 can then be extract from MediaType
  mp3Assert( wmSyncReader->QueryInterface(&wmProfile) );
  mp3Assert( wmProfile->GetStream(0, &wmStreamConfig) );
  mp3Assert( wmStreamConfig->QueryInterface(&wmMediaProperties) );

  // Retrieve sizeof MediaType
  mp3Assert( wmMediaProperties->GetMediaType(NULL, &sizeMediaType) );

  // Retrieve MediaType
  WM_MEDIA_TYPE* mediaType = (WM_MEDIA_TYPE*)LocalAlloc(LPTR,sizeMediaType); 
  mp3Assert( wmMediaProperties->GetMediaType(mediaType, &sizeMediaType) );

  // Check that MediaType is audio
  assert(mediaType->majortype == WMMEDIATYPE_Audio);
  // assert(mediaType->pbFormat == WMFORMAT_WaveFormatEx);

  // Check that input is mp3
  WAVEFORMATEX* inputFormat = (WAVEFORMATEX*)mediaType->pbFormat;
  assert( inputFormat->wFormatTag == WAVE_FORMAT_MPEGLAYER3);
  assert( inputFormat->nSamplesPerSec == 44100);
  assert( inputFormat->nChannels == 2);

  // Release COM interface
  // wmSyncReader->Close();
  wmMediaProperties->Release();
  wmStreamConfig->Release();
  wmProfile->Release();
  wmHeaderInfo->Release();
  wmSyncReader->Release();

  // Free allocated mem
  LocalFree(mediaType);

  // -----------------------------------------------------------------------------------
  // Convert mp3 to pcm using acm driver
  // The following code is mainly inspired from http://david.weekly.org/code/mp3acm.html
  // -----------------------------------------------------------------------------------

  // Get maximum FormatSize for all acm
  mp3Assert( acmMetrics( NULL, ACM_METRIC_MAX_SIZE_FORMAT, &maxFormatSize ) );

  // Allocate PCM output sound buffer
  bufferLength = durationInSecond * pcmFormat.nAvgBytesPerSec;
  soundBuffer = (BYTE*)LocalAlloc(LPTR, bufferLength);

  acmMp3stream = NULL;
  switch ( acmStreamOpen( &acmMp3stream,    // Open an ACM conversion stream
   NULL,                       // Query all ACM drivers
   (LPWAVEFORMATEX)&mp3Format, // input format :  mp3
   &pcmFormat,                 // output format : pcm
   NULL,                       // No filters
   0,                          // No async callback
   0,                          // No data for callback
   0                           // No flags
   ) 
   ) {
      case MMSYSERR_NOERROR:
       break; // success!
      case MMSYSERR_INVALPARAM:
       assert( !"Invalid parameters passed to acmStreamOpen" );
       return E_FAIL;
      case ACMERR_NOTPOSSIBLE:
       assert( !"No ACM filter found capable of decoding MP3" );
       return E_FAIL;
      default:
       assert( !"Some error opening ACM decoding stream!" );
       return E_FAIL;
  }

  // Determine output decompressed buffer size
  unsigned long rawbufsize = 0;
  mp3Assert( acmStreamSize( acmMp3stream, MP3_BLOCK_SIZE, &rawbufsize, ACM_STREAMSIZEF_SOURCE ) );
  assert( rawbufsize > 0 );

  // allocate our I/O buffers
  static BYTE mp3BlockBuffer[MP3_BLOCK_SIZE];
  //LPBYTE mp3BlockBuffer = (LPBYTE) LocalAlloc( LPTR, MP3_BLOCK_SIZE );
  LPBYTE rawbuf = (LPBYTE) LocalAlloc( LPTR, rawbufsize );

  // prepare the decoder
  static ACMSTREAMHEADER mp3streamHead;
  // memset( &mp3streamHead, 0, sizeof(ACMSTREAMHEADER ) );
  mp3streamHead.cbStruct = sizeof(ACMSTREAMHEADER );
  mp3streamHead.pbSrc = mp3BlockBuffer;
  mp3streamHead.cbSrcLength = MP3_BLOCK_SIZE;
  mp3streamHead.pbDst = rawbuf;
  mp3streamHead.cbDstLength = rawbufsize;
  mp3Assert( acmStreamPrepareHeader( acmMp3stream, &mp3streamHead, 0 ) );

  BYTE* currentOutput = soundBuffer;
  DWORD totalDecompressedSize = 0;

  static ULARGE_INTEGER newPosition;
  static LARGE_INTEGER seekValue;
  mp3Assert( mp3Stream->Seek(seekValue, STREAM_SEEK_SET, &newPosition) );

  while(1) {
   // suck in some MP3 data
   ULONG count;
   mp3Assert( mp3Stream->Read(mp3BlockBuffer, MP3_BLOCK_SIZE, &count) );
   if( count != MP3_BLOCK_SIZE )
    break;

   // convert the data
   mp3Assert( acmStreamConvert( acmMp3stream, &mp3streamHead, ACM_STREAMCONVERTF_BLOCKALIGN ) );

   // write the decoded PCM to disk
   //count = fwrite( rawbuf, 1, mp3streamHead.cbDstLengthUsed, fpOut );
   memcpy(currentOutput, rawbuf, mp3streamHead.cbDstLengthUsed);
   totalDecompressedSize += mp3streamHead.cbDstLengthUsed;
   currentOutput += mp3streamHead.cbDstLengthUsed;
  };

  mp3Assert( acmStreamUnprepareHeader( acmMp3stream, &mp3streamHead, 0 ) );
  LocalFree(rawbuf);
  mp3Assert( acmStreamClose( acmMp3stream, 0 ) );

  // Release allocated memory
  mp3Stream->Release();
  GlobalFree(mp3HGlobal);
  return S_OK;
 }

 /*
  * Close : close the current MP3Player, stop playback and free allocated memory
  */
 void __inline Close() {
  // Reset before close (otherwise, waveOutClose will not work on playing buffer)
  waveOutReset(hWaveOut);
  // Close the waveOut
  waveOutClose(hWaveOut);
  // Free allocated memory
  LocalFree(soundBuffer);
 }
 
 /*
  * GetDuration : return the music duration in seconds
  */
 double __inline GetDuration() {
  return durationInSecond;
 }

 /*
  * GetPosition : return the current position from the sound playback (used from sync)
  */
 double GetPosition() {
  static MMTIME MMTime = { TIME_SAMPLES, 0};
  waveOutGetPosition(hWaveOut, &MMTime, sizeof(MMTIME));
  return ((double)MMTime.u.sample)/( 44100.0);
 }

 /*
  * Play : play the previously opened mp3
  */
 void Play() {
  static WAVEHDR WaveHDR = { (LPSTR)soundBuffer,  bufferLength };

  // Define output format
  static WAVEFORMATEX pcmFormat = {
   WAVE_FORMAT_PCM, // WORD        wFormatTag;         /* format type */
   2,     // WORD        nChannels;          /* number of channels (i.e. mono, stereo...) */
   44100,    // DWORD       nSamplesPerSec;     /* sample rate */
   4 * 44100,   // DWORD       nAvgBytesPerSec;    /* for buffer estimation */
   4,     // WORD        nBlockAlign;        /* block size of data */
   16,     // WORD        wBitsPerSample;     /* number of bits per sample of mono data */
   0,     // WORD        cbSize;             /* the count in bytes of the size of */
  };

  mp3Assert( waveOutOpen( &hWaveOut, WAVE_MAPPER, &pcmFormat, NULL, 0, CALLBACK_NULL ) );
  mp3Assert( waveOutPrepareHeader( hWaveOut, &WaveHDR, sizeof(WaveHDR) ) );
  mp3Assert( waveOutWrite  ( hWaveOut, &WaveHDR, sizeof(WaveHDR) ) );
 }
};

#pragma function(memset,memcpy,memcmp)

The usage is then pretty simple :

MP3Player player;

 // Open the mp3 from a file...
 player.OpenFromFile("your.mp3");
 // or From a memory location!
 player.OpenFromMemory(ptrToMP3Song, bytesLength);

 player.Play();

 while (...) {
   // Do here your synchro in the demo using
   ...
   double playerPositionInSeconds = player.GetPosition()
   ...
 }
 player.Close();

And that's all! Hope someone will find this useful!

You can download a Visual Studio project using the MP3Player.h class.

15 comments:

  1. Hello, I think google code is blocking zip files. Have to give them a different extension (".renametozip" perhaps)?

    ReplyDelete
  2. Don't know why, but google groups is screwing up my files... Anyway, link is updated and should work now.

    ReplyDelete
  3. Alright, I have a question. I am a super noob at programming. I just started programming in a c++ programming class, but he's not going to or hasn't taught us yet how to shove mp3s into our program. So, could you help me out by perhaps telling me how you do that? And what be the easiest way? And if not, could you perhaps direct me to a link that would tell me?

    I'm sorry and thank you.

    ReplyDelete
  4. I always thougt myself: "Everything neccesary is here, soooo - there must be a way to do it."
    Long story short: I never figured out how. ;)

    Thanks for sharing your knowledge! :)))

    ReplyDelete
  5. Hi, nice code! I will try to convert the code to handle a mp3 stream instead of a file. Do you have any ideas about how to accomplish that?

    ReplyDelete
  6. @Peter, you have to feed the function acmStreamConvert with a chunk of MP3 buffer (with a size of MP3_BLOCK_SIZE) and send the result to a dynamic audio buffer Instead of getting the data from IStream (mp3Stream in the code), you will get the data from you RTP stream. You will probably have to bufferize the RTP stream, as It will go probably faster than the audio (but you will have also to handle the opposite case). The code above is decoding chunk of MP3 so it is straight transposable to your scenario.

    ReplyDelete
  7. Help me waveOutSetVolume and CALLBACK a file MP3

    ReplyDelete
  8. @Vuong, sorry, if you have problems using the Windows API, you need to ask your question directly to a windows developer forum.

    ReplyDelete
  9. Why this doesn't work with MFC?
    What should I change/add if I want use this amazing class in my project?
    Because when I try to compile this in my MFC project, there are terrible compile-time error. Thanks

    ReplyDelete
  10. Still Works great Alexandre. How would I get this to Loop at the end
    of the song or 1-2 seconds after it ended?

    ReplyDelete
  11. I'm trying to Loop the demo mp3 sound.

    The effect I'm trying to get is Background Music while the user is using the program, so it will play until the user closes the prompt. Any ideas on how I can edit the files to achieve this?

    A play.Loop where I can call it in any file to continuously play any background sound (or close the previous one and loop this one).

    ReplyDelete
  12. Superb work.....really helped me out....keep up

    ReplyDelete
  13. i cant get it to work with a c++ win32 application on microsoft visual studios 2013. i am trying to play a file from my computer files. help please!

    ReplyDelete

Comments are disabled

Note: Only a member of this blog may post a comment.