参考文档

COM Coding Practices
Audio File Format Specifications
Core Audio APIs
Loopback Recording

#include <iostream>
#include <fstream>
#include <vector>

#include <mmdeviceapi.h>
#include <combaseapi.h>
#include <atlbase.h>
#include <Functiondiscoverykeys_devpkey.h>
#include <Audioclient.h>
#include <Audiopolicy.h>

// 利用RAII手法，自动调用 CoUninitialize
class CoInitializeGuard {
public:
    CoInitializeGuard()
    {
        _hr = CoInitializeEx(nullptr, COINIT::COINIT_MULTITHREADED);
    }

    ~CoInitializeGuard()
    {
        if (_hr == S_OK || _hr == S_FALSE) {
            CoUninitialize();
        }
    }

    HRESULT result() const { return _hr; }

private:
    HRESULT _hr;
};

constexpr inline void exit_on_failed(HRESULT hr);
void printEndpoints(CComPtr<IMMDeviceCollection> pColletion);
std::string wchars_to_mbs(const wchar_t* s);

int main()
{
    HRESULT hr{};

    CoInitializeGuard coInitializeGuard;
    exit_on_failed(coInitializeGuard.result());

    // COM 对象都用 CComPtr 包装，会自动调用 Release
    // COM 接口分配的堆变量用 CComHeapPtr 包装，会自动调用 CoTaskMemFree
    CComPtr<IMMDeviceEnumerator> pEnumerator;
    hr = pEnumerator.CoCreateInstance(__uuidof(MMDeviceEnumerator));
    exit_on_failed(hr);

    // 打印所有可用的音频设备
    //CComPtr<IMMDeviceCollection> pColletion;
    //hr = pEnumerator->EnumAudioEndpoints(eRender, DEVICE_STATE_ACTIVE, &pColletion);
    //exit_on_failed(hr);
    //printEndpoints(pColletion);

    // 使用默认的 Audio Endpoint，eRender 表示音频播放设备，而不是录音设备
    CComPtr<IMMDevice> pEndpoint;
    hr = pEnumerator->GetDefaultAudioEndpoint(eRender, eConsole, &pEndpoint);
    exit_on_failed(hr);

    // 打印出播放设备的名字，可能包含中文
    CComPtr<IPropertyStore> pProps;
    hr = pEndpoint->OpenPropertyStore(STGM_READ, &pProps);
    exit_on_failed(hr);
    PROPVARIANT varName;
    PropVariantInit(&varName);
    hr = pProps->GetValue(PKEY_Device_FriendlyName, &varName);
    exit_on_failed(hr);
    std::cout << "select audio endpoint: " << wchars_to_mbs(varName.pwszVal) << std::endl;
    PropVariantClear(&varName);

    // 由 IMMDevice 对象 得到 IAudioClient 对象 
    CComPtr<IAudioClient> pAudioClient;
    hr = pEndpoint->Activate(__uuidof(IAudioClient), CLSCTX_ALL, nullptr, (void**)&pAudioClient);
    exit_on_failed(hr);

    // 获得音频播放设备格式信息
    CComHeapPtr<WAVEFORMATEX> pDeviceFormat;
    pAudioClient->GetMixFormat(&pDeviceFormat);

    constexpr int REFTIMES_PER_SEC = 10000000;      // 1 reference_time = 100ns
    constexpr int REFTIMES_PER_MILLISEC = 10000;

    // 初始化 IAudioClient 对象
    const REFERENCE_TIME hnsRequestedDuration = 2 * REFTIMES_PER_SEC; // 1s
    hr = pAudioClient->Initialize(AUDCLNT_SHAREMODE_SHARED, AUDCLNT_STREAMFLAGS_LOOPBACK, hnsRequestedDuration, 0, pDeviceFormat, nullptr);
    exit_on_failed(hr);

    // 获得缓冲区大小
    UINT32 bufferFrameCount{};
    hr = pAudioClient->GetBufferSize(&bufferFrameCount);
    exit_on_failed(hr);

    // 由 IAudioClient 对象 得到 IAudioCaptureClient 对象，也就是将音频播放设备视为录音设备
    CComPtr<IAudioCaptureClient> pCaptureClient;
    hr = pAudioClient->GetService(__uuidof(IAudioCaptureClient), (void**)&pCaptureClient);
    exit_on_failed(hr);
    
    // 开始录音
    hr = pAudioClient->Start();
    exit_on_failed(hr);

    const REFERENCE_TIME hnsActualDuration = (long long)REFTIMES_PER_SEC * bufferFra