How to increase mp3 decoding quality (Media Foundation)? - c++

Question

I have file .wav that I need to convert in .mp3 in order to do it I am using MediaFoundation. This is approach that I use:
#include "TV_AudioEncoderMF.h"
#include <windows.h>
#include <windowsx.h>
#include <atlstr.h>
#include <comdef.h>
#include <exception>
#include <mfapi.h>
#include <mfplay.h>
#include <mfreadwrite.h>
#include <mmdeviceapi.h>
#include <Audioclient.h>
#include <mferror.h>
#include <Wmcodecdsp.h>
#pragma comment(lib, "mf.lib")
#pragma comment(lib, "mfplat.lib")
#pragma comment(lib, "mfplay.lib")
#pragma comment(lib, "mfreadwrite.lib")
#pragma comment(lib, "mfuuid.lib")
#pragma comment(lib, "wmcodecdspuuid")
TV_AudioEncoderMF::TV_AudioEncoderMF()
{
}
TV_AudioEncoderMF::~TV_AudioEncoderMF()
{
}
template <class T> void SafeRelease(T **ppT)
{
if (*ppT)
{
(*ppT)->Release();
*ppT = nullptr;
}
}
HRESULT TV_AudioEncoderMF::GetOutputMediaTypes(
GUID cAudioFormat,
UINT32 cSampleRate,
UINT32 cBitPerSample,
UINT32 cChannels,
IMFMediaType **ppType
)
{
// Enumerate all codecs except for codecs with field-of-use restrictions.
// Sort the results.
DWORD dwFlags =
(MFT_ENUM_FLAG_ALL & (~MFT_ENUM_FLAG_FIELDOFUSE)) |
MFT_ENUM_FLAG_SORTANDFILTER;
IMFCollection *pAvailableTypes = NULL; // List of audio media types.
IMFMediaType *pAudioType = NULL; // Corresponding codec.
HRESULT hr = MFTranscodeGetAudioOutputAvailableTypes(
cAudioFormat,
dwFlags,
NULL,
&pAvailableTypes
);
// Get the element count.
DWORD dwMTCount;
hr = pAvailableTypes->GetElementCount(&dwMTCount);
// Iterate through the results and check for the corresponding codec.
for (DWORD i = 0; i < dwMTCount; i++)
{
hr = pAvailableTypes->GetElement(i, (IUnknown**)&pAudioType);
GUID majorType;
hr = pAudioType->GetMajorType(&majorType);
GUID subType;
hr = pAudioType->GetGUID(MF_MT_SUBTYPE, &subType);
if (majorType != MFMediaType_Audio || subType != MFAudioFormat_FLAC)
{
continue;
}
UINT32 sampleRate = NULL;
hr = pAudioType->GetUINT32(
MF_MT_AUDIO_SAMPLES_PER_SECOND,
&sampleRate
);
UINT32 bitRate = NULL;
hr = pAudioType->GetUINT32(
MF_MT_AUDIO_BITS_PER_SAMPLE,
&bitRate
);
UINT32 channels = NULL;
hr = pAudioType->GetUINT32(
MF_MT_AUDIO_NUM_CHANNELS,
&channels
);
if (sampleRate == cSampleRate
&& bitRate == cBitPerSample
&& channels == cChannels)
{
// Found the codec.
// Jump out!
break;
}
}
// Add the media type to the caller
*ppType = pAudioType;
(*ppType)->AddRef();
SafeRelease(&pAudioType);
return hr;
}
void TV_AudioEncoderMF::decode()
{
HRESULT hr = S_OK;
// Initialize com interface
CoInitializeEx(0, COINIT_MULTITHREADED);
// Start media foundation
MFStartup(MF_VERSION);
IMFMediaType *pInputType = NULL;
IMFSourceReader *pSourceReader = NULL;
IMFMediaType *pOuputMediaType = NULL;
IMFSinkWriter *pSinkWriter = NULL;
// Create source reader
hr = MFCreateSourceReaderFromURL(
L"D:\\buffer\\del\\out\\test.wav",
NULL,
&pSourceReader
);
// Create sink writer
hr = MFCreateSinkWriterFromURL(
L"D:\\buffer\\del\\out\\test_out.mp3",
NULL,
NULL,
&pSinkWriter
);
// Get media type from source reader
hr = pSourceReader->GetCurrentMediaType(
MF_SOURCE_READER_FIRST_AUDIO_STREAM,
&pInputType
);
// Get sample rate, bit rate and channels
UINT32 sampleRate = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_SAMPLES_PER_SECOND,
&sampleRate
);
UINT32 bitRate = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_BITS_PER_SAMPLE,
&bitRate
);
UINT32 channels = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_NUM_CHANNELS,
&channels
);
// Try to find a media type that is fitting.
hr = GetOutputMediaTypes(
MFAudioFormat_MP3,
sampleRate,
bitRate,
channels,
&pOuputMediaType);
DWORD dwWriterStreamIndex = -1;
// Add the stream
hr = pSinkWriter->AddStream(
pOuputMediaType,
&dwWriterStreamIndex
);
// Set input media type
hr = pSinkWriter->SetInputMediaType(
dwWriterStreamIndex,
pInputType,
NULL
);
// Tell the sink writer to accept data
hr = pSinkWriter->BeginWriting();
// Forever alone loop
while (true)
{
DWORD nStreamIndex, nStreamFlags;
LONGLONG nTime;
IMFSample *pSample;
// Read through the samples until...
hr = pSourceReader->ReadSample(
MF_SOURCE_READER_FIRST_AUDIO_STREAM,
0,
&nStreamIndex,
&nStreamFlags,
&nTime,
&pSample);
if (pSample)
{
hr = pSinkWriter->WriteSample(
dwWriterStreamIndex,
pSample
);
}
// ... we are at the end of the stream...
if (nStreamFlags & MF_SOURCE_READERF_ENDOFSTREAM)
{
// ... and jump out.
break;
}
}
// Call finalize to finish writing.
hr = pSinkWriter->Finalize();
// Done :D
}
Problem is - it is a big difference in audio quality, when I playback (by win standard players) .wav file it sounds good, but when I playback compressed to .mp3 file sound it sounds like person recorded his voice on recorder with a very bad quality.
What is a possible problem here? I don't see any possible way to set out quality, like setOutQualityInPersent(100)
EDIT
void co_AudioEncoderMF::decode()
{
HRESULT hr = S_OK;
// Initialize com interface
CoInitializeEx(0, COINIT_MULTITHREADED);
// Start media foundation
MFStartup(MF_VERSION);
IMFMediaType *pInputType = NULL;
IMFSourceReader *pSourceReader = NULL;
IMFMediaType *pOuputMediaType = NULL;
IMFSinkWriter *pSinkWriter = NULL;
// Create source reader
hr = MFCreateSourceReaderFromURL(
L"D:\\buffer\\del\\out\\test.wav",
NULL,
&pSourceReader
);
// Create sink writer
hr = MFCreateSinkWriterFromURL(
L"D:\\buffer\\del\\out\\test_out.mp3",
NULL,
NULL,
&pSinkWriter
);
// Get media type from source reader
hr = pSourceReader->GetCurrentMediaType(
MF_SOURCE_READER_FIRST_AUDIO_STREAM,
&pInputType
);
// Get sample rate, bit rate and channels
UINT32 sampleRate = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_SAMPLES_PER_SECOND,
&sampleRate
);
UINT32 bitRate = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_BITS_PER_SAMPLE,
&bitRate
);
UINT32 channels = NULL;
hr = pInputType->GetUINT32(
MF_MT_AUDIO_NUM_CHANNELS,
&channels
);
// Try to find a media type that is fitting.
hr = GetOutputMediaTypes(
MFAudioFormat_MP3,
sampleRate,
bitRate,
channels,
&pOuputMediaType);
bitRate = bitRate + 2; <------- This line
pOuputMediaType->SetUINT32(MF_MT_AUDIO_BITS_PER_SAMPLE, bitRate); <------- This line
DWORD dwWriterStreamIndex = -1;
// Add the stream
hr = pSinkWriter->AddStream(
pOuputMediaType,
&dwWriterStreamIndex
);
// Set input media type
hr = pSinkWriter->SetInputMediaType(
dwWriterStreamIndex,
pInputType,
NULL
);
// Tell the sink writer to accept data
hr = pSinkWriter->BeginWriting();
// Forever alone loop
while (true)
{
DWORD nStreamIndex, nStreamFlags;
LONGLONG nTime;
IMFSample *pSample;
// Read through the samples until...
hr = pSourceReader->ReadSample(
MF_SOURCE_READER_FIRST_AUDIO_STREAM,
0,
&nStreamIndex,
&nStreamFlags,
&nTime,
&pSample);
if (pSample)
{
hr = pSinkWriter->WriteSample(
dwWriterStreamIndex,
pSample
);
}
// ... we are at the end of the stream...
if (nStreamFlags & MF_SOURCE_READERF_ENDOFSTREAM)
{
// ... and jump out.
break;
}
}
// Call finalize to finish writing.
hr = pSinkWriter->Finalize();
// Done :D
}
EDIT2
There are 2 files - https://drive.google.com/drive/folders/1yzB2u0TvMSnwsTpYnDDPFBDkTB75ZFwM?usp=sharing
Result and orig

Answer 1

This part is just broken:
// Try to find a media type that is fitting.
hr = GetOutputMediaTypes(
MFAudioFormat_MP3,
sampleRate,
bitRate,
channels,
&pOuputMediaType);
bitRate = bitRate + 2; <------- This line
pOuputMediaType->SetUINT32(MF_MT_AUDIO_BITS_PER_SAMPLE, bitRate); <------- This line
To get you back on track, replace the fragment above with:
MFCreateMediaType(&pOuputMediaType);
pOuputMediaType->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Audio);
pOuputMediaType->SetGUID(MF_MT_SUBTYPE, MFAudioFormat_MP3);
pOuputMediaType->SetUINT32(MF_MT_AUDIO_AVG_BYTES_PER_SECOND, 128000 / 8);
pOuputMediaType->SetUINT32(MF_MT_AUDIO_NUM_CHANNELS, channels);
pOuputMediaType->SetUINT32(MF_MT_AUDIO_SAMPLES_PER_SECOND, sampleRate);
and you'll start getting proper MP3.
Note that the attributes above are taken directly from documentation: MP3 Audio Encoder. In your application you will need to make sure that target values remain valid and match the documented options. You might need to resample audio, for example.