音频特效滤镜 via Media Foundation Transform (MFT)

音频特效定义

音效或声效（Sound effects 或 Audio effects）是人工制造或加强的声音，用来增强对电影、电子游戏、音乐或其他媒体的艺术或其他内容的声音处理。

常见的音效技术有：回声、合唱、均衡（EQ）、过滤、变调、移相、压缩 / 拉伸、3D、调制和共鸣等等。

Media Foundation Transform

MF 中插件是以 MFT 的形式创建的，需要继承 IMFTransform 接口，接口函数如下：

// Methods That Handle Format Negotiation
STDMETHODIMP GetStreamLimits(DWORD *pInputMinimum, DWORD *pInputMaximum, DWORD *pOutputMinimum, DWORD *pOutputMaximum);
STDMETHODIMP GetStreamCount(DWORD *pcInputStreams, DWORD *pcOutputStreams);
STDMETHODIMP GetStreamIDs(DWORD dwInputIDArraySize, DWORD *pdwInputIDs, DWORD dwOutputIDArraySize, DWORD *pdwOutputIDs);
STDMETHODIMP GetInputStreamInfo(DWORD dwInputStreamID, MFT_INPUT_STREAM_INFO *pStreamInfo);
STDMETHODIMP GetOutputStreamInfo(DWORD dwOutputStreamID, MFT_OUTPUT_STREAM_INFO *pStreamInfo);
STDMETHODIMP GetAttributes(IMFAttributes **ppAttributes);
STDMETHODIMP GetInputStreamAttributes(DWORD dwInputStreamID, IMFAttributes **ppAttributes);
STDMETHODIMP GetOutputStreamAttributes(DWORD dwOutputStreamID, IMFAttributes **ppAttributes);
STDMETHODIMP GetInputAvailableType(DWORD dwInputStreamID, DWORD dwTypeIndex, IMFMediaType **ppType);
STDMETHODIMP GetOutputAvailableType(DWORD dwOutputStreamID, DWORD dwTypeIndex, IMFMediaType **ppType);
STDMETHODIMP SetInputType(DWORD dwInputStreamID, IMFMediaType *pType, DWORD dwFlags);
STDMETHODIMP SetOutputType(DWORD dwOutputStreamID, IMFMediaType *pType, DWORD dwFlags);

// Methods That Specify or Retrieve State Information
STDMETHODIMP GetInputCurrentType(DWORD dwInputStreamID, IMFMediaType **ppType);
STDMETHODIMP GetOutputCurrentType(DWORD dwOutputStreamID, IMFMediaType **ppType);
STDMETHODIMP DeleteInputStream(DWORD dwStreamID);
STDMETHODIMP AddInputStreams(DWORD cStreams, DWORD *adwStreamIDs);
STDMETHODIMP GetInputStatus(DWORD dwInputStreamID, DWORD *pdwFlags);
STDMETHODIMP GetOutputStatus(DWORD *pdwFlags);
STDMETHODIMP SetOutputBounds(LONGLONG hnsLowerBound, LONGLONG hnsUpperBound);

// Methods That Handle Buffering and Processing Data
STDMETHODIMP ProcessEvent(DWORD dwInputStreamID, IMFMediaEvent *pEvent);
STDMETHODIMP ProcessMessage(MFT_MESSAGE_TYPE eMessage, ULONG_PTR ulParam);
STDMETHODIMP ProcessInput(DWORD dwInputStreamID, IMFSample *pSample, DWORD dwFlags);
STDMETHODIMP ProcessOutput(DWORD dwFlags, DWORD cOutputBufferCount, MFT_OUTPUT_DATA_BUFFER *pOutputSamples, DWORD *pStatus);

如果要实现一个 异步的 MFT，则还需要继承 IMFMediaEventGenerator 和 IMFShutdown 接口。

// IMFMediaEventGenerator
HRESULT BeginGetEvent(IMFAsyncCallback *pCallback, IUnknown *punkState);
HRESULT EndGetEvent(IMFAsyncResult *pResult, IMFMediaEvent **ppEvent);
HRESULT GetEvent(DWORD dwFlags, IMFMediaEvent **ppEvent);
HRESULT QueueEvent(MediaEventType met, REFGUID guidExtendedType, HRESULT hrStatus, const PROPVARIANT *pvValue);

// IMFShutdown
HRESULT GetShutdownStatus(MFSHUTDOWN_STATUS *pStatus);
HRESULT Shutdown();

下面介绍一下 IMFTransform 的几个比较重要的接口，以一个延迟音频的插件为例。
代码包含在 Windows SDK 7.x 的 samples\multimedia\mediafoundation\mft_audiodelay\ 目录下。

IMFTransform::GetInputStreamInfo 函数

对输入流的要求，比如音频 sample 必须完整且大小固定，可以 in place 直接处理，无需再分配内存，等等。

HRESULT CDelayMFT::GetInputStreamInfo(DWORD dwInputStreamID, MFT_INPUT_STREAM_INFO *pStreamInfo)
{
   
   
    AutoLock lock(m_critSec);    
    if (!IsValidInputStream(dwInputStreamID))
        return MF_E_INVALIDSTREAMNUMBER;

    pStreamInfo->hnsMaxLatency = 0;
    pStreamInfo->dwFlags = 
        MFT_INPUT_STREAM_WHOLE_SAMPLES |        // The MFT must get complete audio frames. 
        MFT_INPUT_STREAM_PROCESSES_IN_PLACE |   // The MFT can do in-place processing.
        MFT_INPUT_STREAM_FIXED_SAMPLE_SIZE;     // Samples (i.e., audio frames) are fixed size.

    pStreamInfo->cbSize = 0;  // If no media type is set, use zero. 
    pStreamInfo->cbMaxLookahead = 0;
    pStreamInfo->cbAlignment = 0;

    // When the media type is set, return the minimum buffer size = one audio frame.
    if (IsInputTypeSet())
        pStreamInfo->cbSize = BlockAlign();

    return S_OK;
}

IMFTransform::GetOutputStreamInfo 函数

输出流的信息，比如音频 sample 是完整且固定大小的，可使用 sample 原来分配的 buffer 或按需自己分配，等等。

HRESULT CDelayMFT::GetOutputStreamInfo(DWORD dwOutputStreamID, MFT_OUTPUT_STREAM_INFO *pStreamInfo)
{
   
   
    AutoLock lock(m_critSec);

    if (!IsValidOutputStream(dwOutputStreamID))
        return MF_E_INVALIDSTREAMNUMBER;

    pStreamInfo->dwFlags = 
        MFT_OUTPUT_STREAM_WHOLE_SAMPLES |         // Output buffers contain complete audio frames. 
        MFT_OUTPUT_STREAM_CAN_PROVIDE_SAMPLES |   // The MFT can allocate output buffers, 
                                                  // or use caller-allocated buffers.