Use factory method to create screen capturer

src/screen_capturer/linux/screen_capturer_x11.cpp

@@ -0,0 +1,143 @@
+#include "screen_capturer_x11.h"
+
+#include <iostream>
+
+#include "log.h"
+
+#define NV12_BUFFER_SIZE 1280 * 720 * 3 / 2
+unsigned char nv12_buffer_[NV12_BUFFER_SIZE];
+
+ScreenCapturerX11::ScreenCapturerX11() {}
+
+ScreenCapturerX11::~ScreenCapturerX11() {}
+
+int ScreenCapturerX11::Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                            cb_desktop_data cb) {
+  if (cb) {
+    _on_data = cb;
+  }
+
+  fps_ = fps;
+
+  av_log_set_level(AV_LOG_QUIET);
+
+  pFormatCtx_ = avformat_alloc_context();
+
+  avdevice_register_all();
+
+  // grabbing frame rate
+  av_dict_set(&options_, "framerate", "30", 0);
+  // Make the grabbed area follow the mouse
+  // av_dict_set(&options_, "follow_mouse", "centered", 0);
+  // Video frame size. The default is to capture the full screen
+  // av_dict_set(&options_, "video_size", "1280x720", 0);
+  std::string capture_method = "x11grab";
+  ifmt_ = (AVInputFormat *)av_find_input_format(capture_method.c_str());
+  if (!ifmt_) {
+    LOG_ERROR("Couldn't find_input_format [{}]", capture_method.c_str());
+  }
+
+  // Grab at position 10,20
+  if (avformat_open_input(&pFormatCtx_, ":0.0", ifmt_, &options_) != 0) {
+    printf("Couldn't open input stream.\n");
+    return -1;
+  }
+
+  if (avformat_find_stream_info(pFormatCtx_, NULL) < 0) {
+    printf("Couldn't find stream information.\n");
+    return -1;
+  }
+
+  videoindex_ = -1;
+  for (i_ = 0; i_ < pFormatCtx_->nb_streams; i_++)
+    if (pFormatCtx_->streams[i_]->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
+      videoindex_ = i_;
+      break;
+    }
+  if (videoindex_ == -1) {
+    printf("Didn't find a video stream.\n");
+    return -1;
+  }
+
+  pCodecParam_ = pFormatCtx_->streams[videoindex_]->codecpar;
+
+  pCodecCtx_ = avcodec_alloc_context3(NULL);
+  avcodec_parameters_to_context(pCodecCtx_, pCodecParam_);
+
+  pCodec_ = const_cast<AVCodec *>(avcodec_find_decoder(pCodecCtx_->codec_id));
+  if (pCodec_ == NULL) {
+    printf("Codec not found.\n");
+    return -1;
+  }
+  if (avcodec_open2(pCodecCtx_, pCodec_, NULL) < 0) {
+    printf("Could not open codec.\n");
+    return -1;
+  }
+
+  const int screen_w = pFormatCtx_->streams[videoindex_]->codecpar->width;
+  const int screen_h = pFormatCtx_->streams[videoindex_]->codecpar->height;
+
+  pFrame_ = av_frame_alloc();
+  pFrameNv12_ = av_frame_alloc();
+
+  pFrame_->width = screen_w;
+  pFrame_->height = screen_h;
+  pFrameNv12_->width = 1280;
+  pFrameNv12_->height = 720;
+
+  packet_ = (AVPacket *)av_malloc(sizeof(AVPacket));
+
+  img_convert_ctx_ = sws_getContext(
+      pFrame_->width, pFrame_->height, pCodecCtx_->pix_fmt, pFrameNv12_->width,
+      pFrameNv12_->height, AV_PIX_FMT_NV12, SWS_BICUBIC, NULL, NULL, NULL);
+
+  return 0;
+}
+
+int ScreenCapturerX11::Destroy() {
+  if (capture_thread_->joinable()) {
+    capture_thread_->join();
+  }
+
+  return 0;
+}
+
+int ScreenCapturerX11::Start() {
+  capture_thread_.reset(new std::thread([this]() {
+    while (1) {
+      if (av_read_frame(pFormatCtx_, packet_) >= 0) {
+        if (packet_->stream_index == videoindex_) {
+          avcodec_send_packet(pCodecCtx_, packet_);
+          av_packet_unref(packet_);
+          got_picture_ = avcodec_receive_frame(pCodecCtx_, pFrame_);
+
+          if (!got_picture_) {
+            av_image_fill_arrays(pFrameNv12_->data, pFrameNv12_->linesize,
+                                 nv12_buffer_, AV_PIX_FMT_NV12,
+                                 pFrameNv12_->width, pFrameNv12_->height, 1);
+
+            sws_scale(img_convert_ctx_, pFrame_->data, pFrame_->linesize, 0,
+                      pFrame_->height, pFrameNv12_->data,
+                      pFrameNv12_->linesize);
+
+            _on_data((unsigned char *)nv12_buffer_,
+                     pFrameNv12_->width * pFrameNv12_->height * 3 / 2,
+                     pFrameNv12_->width, pFrameNv12_->height);
+          }
+        }
+      }
+    }
+  }));
+
+  return 0;
+}
+
+int ScreenCapturerX11::Pause() { return 0; }
+
+int ScreenCapturerX11::Resume() { return 0; }
+
+int ScreenCapturerX11::Stop() { return 0; }
+
+void ScreenCapturerX11::OnFrame() {}
+
+void ScreenCapturerX11::CleanUp() {}

src/screen_capturer/linux/screen_capturer_x11.h

@@ -0,0 +1,80 @@
+#ifndef _SCREEN_CAPTURER_X11_H_
+#define _SCREEN_CAPTURER_X11_H_
+
+#include <atomic>
+#include <functional>
+#include <string>
+#include <thread>
+
+#include "screen_capturer.h"
+#ifdef __cplusplus
+extern "C" {
+#endif
+#include <libavcodec/avcodec.h>
+#include <libavdevice/avdevice.h>
+#include <libavformat/avformat.h>
+#include <libavutil/imgutils.h>
+#include <libswscale/swscale.h>
+#ifdef __cplusplus
+};
+#endif
+
+class ScreenCapturerX11 : public ScreenCapturer {
+ public:
+  ScreenCapturerX11();
+  ~ScreenCapturerX11();
+
+ public:
+  virtual int Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                   cb_desktop_data cb);
+
+  virtual int Destroy();
+
+  virtual int Start();
+
+  int Pause();
+  int Resume();
+  int Stop();
+
+  void OnFrame();
+
+ protected:
+  void CleanUp();
+
+ private:
+  std::atomic_bool _running;
+  std::atomic_bool _paused;
+  std::atomic_bool _inited;
+
+  std::thread _thread;
+
+  std::string _device_name;
+
+  RECORD_DESKTOP_RECT _rect;
+
+  int _fps;
+
+  cb_desktop_data _on_data;
+
+ private:
+  int i_ = 0;
+  int videoindex_ = 0;
+  int got_picture_ = 0;
+  int fps_ = 0;
+  // ffmpeg
+  AVFormatContext *pFormatCtx_ = nullptr;
+  AVCodecContext *pCodecCtx_ = nullptr;
+  AVCodec *pCodec_ = nullptr;
+  AVCodecParameters *pCodecParam_ = nullptr;
+  AVDictionary *options_ = nullptr;
+  AVInputFormat *ifmt_ = nullptr;
+  AVFrame *pFrame_ = nullptr;
+  AVFrame *pFrameNv12_ = nullptr;
+  AVPacket *packet_ = nullptr;
+  struct SwsContext *img_convert_ctx_ = nullptr;
+
+  // thread
+  std::unique_ptr<std::thread> capture_thread_ = nullptr;
+};
+
+#endif

src/screen_capturer/linux/x11_session.h

@@ -0,0 +1,37 @@
+#ifndef _X11_SESSION_H_
+#define _X11_SESSION_H_
+
+class X11Session {
+ public:
+  struct x11_session_frame {
+    unsigned int width;
+    unsigned int height;
+    unsigned int row_pitch;
+
+    const unsigned char *data;
+  };
+
+  class x11_session_observer {
+   public:
+    virtual ~x11_session_observer() {}
+    virtual void OnFrame(const x11_session_frame &frame) = 0;
+  };
+
+ public:
+  virtual void Release() = 0;
+
+  virtual int Initialize() = 0;
+
+  virtual void RegisterObserver(x11_session_observer *observer) = 0;
+
+  virtual int Start() = 0;
+  virtual int Stop() = 0;
+
+  virtual int Pause() = 0;
+  virtual int Resume() = 0;
+
+ protected:
+  virtual ~X11Session(){};
+};
+
+#endif

src/screen_capturer/linux/x11_session_impl.cpp

@@ -0,0 +1,49 @@
+#include "x11_session_impl.h"
+
+#include <atomic>
+#include <functional>
+#include <iostream>
+#include <memory>
+
+#define CHECK_INIT                            \
+  if (!is_initialized_) {                     \
+    std::cout << "AE_NEED_INIT" << std::endl; \
+    return 4;                                 \
+  }
+
+X11SessionImpl::X11SessionImpl() {}
+
+X11SessionImpl::~X11SessionImpl() {
+  Stop();
+  CleanUp();
+}
+
+void X11SessionImpl::Release() { delete this; }
+
+int X11SessionImpl::Initialize() { return 0; }
+
+void X11SessionImpl::RegisterObserver(x11_session_observer *observer) {
+  observer_ = observer;
+}
+
+int X11SessionImpl::Start() {
+  if (is_running_) return 0;
+
+  int error = 1;
+
+  CHECK_INIT;
+
+  return error;
+}
+
+int X11SessionImpl::Stop() { return 0; }
+
+int X11SessionImpl::Pause() { return 0; }
+
+int X11SessionImpl::Resume() { return 0; }
+
+void X11SessionImpl::OnFrame() {}
+
+void X11SessionImpl::OnClosed() {}
+
+void X11SessionImpl::CleanUp() {}

src/screen_capturer/linux/x11_session_impl.h

@@ -0,0 +1,44 @@
+#ifndef _WGC_SESSION_IMPL_H_
+#define _WGC_SESSION_IMPL_H_
+
+#include <mutex>
+#include <thread>
+
+#include "x11_session.h"
+
+class X11SessionImpl : public X11Session {
+ public:
+  X11SessionImpl();
+  ~X11SessionImpl() override;
+
+ public:
+  void Release() override;
+
+  int Initialize() override;
+
+  void RegisterObserver(x11_session_observer *observer) override;
+
+  int Start() override;
+  int Stop() override;
+
+  int Pause() override;
+  int Resume() override;
+
+ private:
+  void OnFrame();
+  void OnClosed();
+
+  void CleanUp();
+
+  // void message_func();
+
+ private:
+  std::mutex lock_;
+  bool is_initialized_ = false;
+  bool is_running_ = false;
+  bool is_paused_ = false;
+
+  x11_session_observer *observer_ = nullptr;
+};
+
+#endif

src/screen_capturer/macosx/screen_capturer_avf.cpp

@@ -0,0 +1,144 @@
+#include "screen_capturer_avf.h"
+
+#include <iostream>
+
+#include "log.h"
+
+#define NV12_BUFFER_SIZE 1280 * 720 * 3 / 2
+unsigned char nv12_buffer_[NV12_BUFFER_SIZE];
+
+ScreenCapturerAvf::ScreenCapturerAvf() {}
+
+ScreenCapturerAvf::~ScreenCapturerAvf() {}
+
+int ScreenCapturerAvf::Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                            cb_desktop_data cb) {
+  if (cb) {
+    _on_data = cb;
+  }
+
+  av_log_set_level(AV_LOG_QUIET);
+
+  pFormatCtx_ = avformat_alloc_context();
+
+  avdevice_register_all();
+
+  // grabbing frame rate
+  av_dict_set(&options_, "framerate", "60", 0);
+  av_dict_set(&options_, "pixel_format", "nv12", 0);
+  // show remote cursor
+  av_dict_set(&options_, "capture_cursor", "1", 0);
+  // Make the grabbed area follow the mouse
+  // av_dict_set(&options_, "follow_mouse", "centered", 0);
+  // Video frame size. The default is to capture the full screen
+  // av_dict_set(&options_, "video_size", "1280x720", 0);
+  ifmt_ = (AVInputFormat *)av_find_input_format("avfoundation");
+  if (!ifmt_) {
+    printf("Couldn't find_input_format\n");
+  }
+
+  // Grab at position 10,20
+  if (avformat_open_input(&pFormatCtx_, "Capture screen 0", ifmt_, &options_) !=
+      0) {
+    printf("Couldn't open input stream.\n");
+    return -1;
+  }
+
+  if (avformat_find_stream_info(pFormatCtx_, NULL) < 0) {
+    printf("Couldn't find stream information.\n");
+    return -1;
+  }
+
+  videoindex_ = -1;
+  for (i_ = 0; i_ < pFormatCtx_->nb_streams; i_++)
+    if (pFormatCtx_->streams[i_]->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
+      videoindex_ = i_;
+      break;
+    }
+  if (videoindex_ == -1) {
+    printf("Didn't find a video stream.\n");
+    return -1;
+  }
+
+  pCodecParam_ = pFormatCtx_->streams[videoindex_]->codecpar;
+
+  pCodecCtx_ = avcodec_alloc_context3(NULL);
+  avcodec_parameters_to_context(pCodecCtx_, pCodecParam_);
+
+  pCodec_ = const_cast<AVCodec *>(avcodec_find_decoder(pCodecCtx_->codec_id));
+  if (pCodec_ == NULL) {
+    printf("Codec not found.\n");
+    return -1;
+  }
+  if (avcodec_open2(pCodecCtx_, pCodec_, NULL) < 0) {
+    printf("Could not open codec.\n");
+    return -1;
+  }
+
+  const int screen_w = pFormatCtx_->streams[videoindex_]->codecpar->width;
+  const int screen_h = pFormatCtx_->streams[videoindex_]->codecpar->height;
+
+  pFrame_ = av_frame_alloc();
+  pFrameNv12_ = av_frame_alloc();
+
+  pFrame_->width = screen_w;
+  pFrame_->height = screen_h;
+  pFrameNv12_->width = 1280;
+  pFrameNv12_->height = 720;
+
+  packet_ = (AVPacket *)av_malloc(sizeof(AVPacket));
+
+  img_convert_ctx_ = sws_getContext(
+      pFrame_->width, pFrame_->height, pCodecCtx_->pix_fmt, pFrameNv12_->width,
+      pFrameNv12_->height, AV_PIX_FMT_NV12, SWS_BICUBIC, NULL, NULL, NULL);
+
+  return 0;
+}
+
+int ScreenCapturerAvf::Destroy() {
+  if (capture_thread_->joinable()) {
+    capture_thread_->join();
+  }
+
+  return 0;
+}
+
+int ScreenCapturerAvf::Start() {
+  capture_thread_.reset(new std::thread([this]() {
+    while (1) {
+      if (av_read_frame(pFormatCtx_, packet_) >= 0) {
+        if (packet_->stream_index == videoindex_) {
+          avcodec_send_packet(pCodecCtx_, packet_);
+          av_packet_unref(packet_);
+          got_picture_ = avcodec_receive_frame(pCodecCtx_, pFrame_);
+
+          if (!got_picture_) {
+            av_image_fill_arrays(pFrameNv12_->data, pFrameNv12_->linesize,
+                                 nv12_buffer_, AV_PIX_FMT_NV12,
+                                 pFrameNv12_->width, pFrameNv12_->height, 1);
+
+            sws_scale(img_convert_ctx_, pFrame_->data, pFrame_->linesize, 0,
+                      pFrame_->height, pFrameNv12_->data,
+                      pFrameNv12_->linesize);
+
+            _on_data((unsigned char *)nv12_buffer_,
+                     pFrameNv12_->width * pFrameNv12_->height * 3 / 2,
+                     pFrameNv12_->width, pFrameNv12_->height);
+          }
+        }
+      }
+    }
+  }));
+
+  return 0;
+}
+
+int ScreenCapturerAvf::Pause() { return 0; }
+
+int ScreenCapturerAvf::Resume() { return 0; }
+
+int ScreenCapturerAvf::Stop() { return 0; }
+
+void ScreenCapturerAvf::OnFrame() {}
+
+void ScreenCapturerAvf::CleanUp() {}

src/screen_capturer/macosx/screen_capturer_avf.h

@@ -0,0 +1,85 @@
+/*
+ * @Author: DI JUNKUN
+ * @Date: 2023-12-01
+ * Copyright (c) 2023 by DI JUNKUN, All Rights Reserved.
+ */
+
+#ifndef _SCREEN_CAPTURER_AVF_H_
+#define _SCREEN_CAPTURER_AVF_H_
+
+#include <atomic>
+#include <functional>
+#include <string>
+#include <thread>
+
+#include "screen_capturer.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+#include <libavcodec/avcodec.h>
+#include <libavdevice/avdevice.h>
+#include <libavformat/avformat.h>
+#include <libavutil/imgutils.h>
+#include <libswscale/swscale.h>
+#ifdef __cplusplus
+};
+#endif
+
+class ScreenCapturerAvf : public ScreenCapturer {
+ public:
+  ScreenCapturerAvf();
+  ~ScreenCapturerAvf();
+
+ public:
+  virtual int Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                   cb_desktop_data cb);
+  virtual int Destroy();
+
+  virtual int Start();
+
+  int Pause();
+  int Resume();
+  int Stop();
+
+  void OnFrame();
+
+ protected:
+  void CleanUp();
+
+ private:
+  std::atomic_bool _running;
+  std::atomic_bool _paused;
+  std::atomic_bool _inited;
+
+  std::thread _thread;
+
+  std::string _device_name;
+
+  RECORD_DESKTOP_RECT _rect;
+
+  int _fps;
+
+  cb_desktop_data _on_data;
+
+ private:
+  int i_ = 0;
+  int videoindex_ = 0;
+  int got_picture_ = 0;
+  // ffmpeg
+  AVFormatContext *pFormatCtx_ = nullptr;
+  AVCodecContext *pCodecCtx_ = nullptr;
+  AVCodec *pCodec_ = nullptr;
+  AVCodecParameters *pCodecParam_ = nullptr;
+  AVDictionary *options_ = nullptr;
+  AVInputFormat *ifmt_ = nullptr;
+  AVFrame *pFrame_ = nullptr;
+  AVFrame *pFrameNv12_ = nullptr;
+  AVPacket *packet_ = nullptr;
+  struct SwsContext *img_convert_ctx_ = nullptr;
+
+  // thread
+  std::unique_ptr<std::thread> capture_thread_ = nullptr;
+};
+
+#endif

src/screen_capturer/screen_capturer.h

@@ -0,0 +1,33 @@
+/*
+ * @Author: DI JUNKUN
+ * @Date: 2023-12-15
+ * Copyright (c) 2023 by DI JUNKUN, All Rights Reserved.
+ */
+
+#ifndef _SCREEN_CAPTURER_H_
+#define _SCREEN_CAPTURER_H_
+
+#include <functional>
+
+class ScreenCapturer {
+ public:
+  typedef struct {
+    int left;
+    int top;
+    int right;
+    int bottom;
+  } RECORD_DESKTOP_RECT;
+  typedef std::function<void(unsigned char *, int, int, int)> cb_desktop_data;
+
+ public:
+  virtual ~ScreenCapturer() {}
+
+ public:
+  virtual int Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                   cb_desktop_data cb) = 0;
+  virtual int Destroy() = 0;
+
+  virtual int Start() = 0;
+};
+
+#endif

src/screen_capturer/screen_capturer_factory.h

@@ -0,0 +1,37 @@
+/*
+ * @Author: DI JUNKUN
+ * @Date: 2023-12-15
+ * Copyright (c) 2023 by DI JUNKUN, All Rights Reserved.
+ */
+
+#ifndef _SCREEN_CAPTURER_FACTORY_H_
+#define _SCREEN_CAPTURER_FACTORY_H_
+
+#ifdef _WIN32
+
+#include "screen_capturer_wgc.h"
+#elif __linux__
+#include "screen_capturer_x11.h"
+#elif __APPLE__
+#include "screen_capturer_avf.h"
+#endif
+
+class ScreenCapturerFactory {
+ public:
+  virtual ~ScreenCapturerFactory() {}
+
+ public:
+  ScreenCapturer* Create() {
+#ifdef _WIN32
+    return new ScreenCapturerWgc();
+#elif __linux__
+    return new ScreenCapturerX11();
+#elif __APPLE__
+    return new ScreenCapturerAvf();
+#else
+    return nullptr;
+#endif
+  }
+};
+
+#endif

src/screen_capturer/windows/screen_capturer_wgc.cpp

@@ -0,0 +1,180 @@
+#include "screen_capturer_wgc.h"
+
+#include <Windows.h>
+#include <d3d11_4.h>
+#include <winrt/Windows.Foundation.Metadata.h>
+#include <winrt/Windows.Graphics.Capture.h>
+
+extern "C" {
+#include <libavutil/imgutils.h>
+#include <libswscale/swscale.h>
+};
+
+#include <iostream>
+
+int BGRAToNV12FFmpeg(unsigned char *src_buffer, int width, int height,
+                     unsigned char *dst_buffer) {
+  AVFrame *Input_pFrame = av_frame_alloc();
+  AVFrame *Output_pFrame = av_frame_alloc();
+  struct SwsContext *img_convert_ctx =
+      sws_getContext(width, height, AV_PIX_FMT_BGRA, 1280, 720, AV_PIX_FMT_NV12,
+                     SWS_FAST_BILINEAR, nullptr, nullptr, nullptr);
+
+  av_image_fill_arrays(Input_pFrame->data, Input_pFrame->linesize, src_buffer,
+                       AV_PIX_FMT_BGRA, width, height, 1);
+  av_image_fill_arrays(Output_pFrame->data, Output_pFrame->linesize, dst_buffer,
+                       AV_PIX_FMT_NV12, 1280, 720, 1);
+
+  sws_scale(img_convert_ctx, (uint8_t const **)Input_pFrame->data,
+            Input_pFrame->linesize, 0, height, Output_pFrame->data,
+            Output_pFrame->linesize);
+
+  if (Input_pFrame) av_free(Input_pFrame);
+  if (Output_pFrame) av_free(Output_pFrame);
+  if (img_convert_ctx) sws_freeContext(img_convert_ctx);
+
+  return 0;
+}
+
+BOOL WINAPI EnumMonitorProc(HMONITOR hmonitor, HDC hdc, LPRECT lprc,
+                            LPARAM data) {
+  MONITORINFOEX info_ex;
+  info_ex.cbSize = sizeof(MONITORINFOEX);
+
+  GetMonitorInfo(hmonitor, &info_ex);
+
+  if (info_ex.dwFlags == DISPLAY_DEVICE_MIRRORING_DRIVER) return true;
+
+  if (info_ex.dwFlags & MONITORINFOF_PRIMARY) {
+    *(HMONITOR *)data = hmonitor;
+  }
+
+  return true;
+}
+
+HMONITOR GetPrimaryMonitor() {
+  HMONITOR hmonitor = nullptr;
+
+  ::EnumDisplayMonitors(NULL, NULL, EnumMonitorProc, (LPARAM)&hmonitor);
+
+  return hmonitor;
+}
+
+ScreenCapturerWgc::ScreenCapturerWgc() {}
+
+ScreenCapturerWgc::~ScreenCapturerWgc() {}
+
+bool ScreenCapturerWgc::IsWgcSupported() {
+  try {
+    /* no contract for IGraphicsCaptureItemInterop, verify 10.0.18362.0 */
+    return winrt::Windows::Foundation::Metadata::ApiInformation::
+        IsApiContractPresent(L"Windows.Foundation.UniversalApiContract", 8);
+  } catch (const winrt::hresult_error &) {
+    return false;
+  } catch (...) {
+    return false;
+  }
+}
+
+int ScreenCapturerWgc::Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                            cb_desktop_data cb) {
+  int error = 0;
+  if (_inited == true) return error;
+
+  nv12_frame_ = new unsigned char[rect.right * rect.bottom * 4];
+
+  _fps = fps;
+
+  _on_data = cb;
+
+  do {
+    if (!IsWgcSupported()) {
+      std::cout << "AE_UNSUPPORT" << std::endl;
+      error = 2;
+      break;
+    }
+
+    session_ = new WgcSessionImpl();
+    if (!session_) {
+      error = -1;
+      std::cout << "AE_WGC_CREATE_CAPTURER_FAILED" << std::endl;
+      break;
+    }
+
+    session_->RegisterObserver(this);
+
+    error = session_->Initialize(GetPrimaryMonitor());
+
+    _inited = true;
+  } while (0);
+
+  if (error != 0) {
+  }
+
+  return error;
+}
+
+int ScreenCapturerWgc::Destroy() {
+  if (nv12_frame_) {
+    delete nv12_frame_;
+    nv12_frame_ = nullptr;
+  }
+
+  Stop();
+  CleanUp();
+
+  return 0;
+}
+
+int ScreenCapturerWgc::Start() {
+  if (_running == true) {
+    std::cout << "record desktop duplication is already running" << std::endl;
+    return 0;
+  }
+
+  if (_inited == false) {
+    std::cout << "AE_NEED_INIT" << std::endl;
+    return 4;
+  }
+
+  _running = true;
+  session_->Start();
+
+  return 0;
+}
+
+int ScreenCapturerWgc::Pause() {
+  _paused = true;
+  if (session_) session_->Pause();
+  return 0;
+}
+
+int ScreenCapturerWgc::Resume() {
+  _paused = false;
+  if (session_) session_->Resume();
+  return 0;
+}
+
+int ScreenCapturerWgc::Stop() {
+  _running = false;
+
+  if (session_) session_->Stop();
+
+  return 0;
+}
+
+void ScreenCapturerWgc::OnFrame(const WgcSession::wgc_session_frame &frame) {
+  if (_on_data)
+    BGRAToNV12FFmpeg((unsigned char *)frame.data, frame.width, frame.height,
+                     nv12_frame_);
+  _on_data(nv12_frame_, frame.width * frame.height * 4, frame.width,
+           frame.height);
+}
+
+void ScreenCapturerWgc::CleanUp() {
+  _inited = false;
+
+  if (session_) session_->Release();
+
+  session_ = nullptr;
+}

src/screen_capturer/windows/screen_capturer_wgc.h

@@ -0,0 +1,57 @@
+#ifndef _SCREEN_CAPTURER_WGC_H_
+#define _SCREEN_CAPTURER_WGC_H_
+
+#include <atomic>
+#include <functional>
+#include <string>
+#include <thread>
+
+#include "screen_capturer.h"
+#include "wgc_session.h"
+#include "wgc_session_impl.h"
+
+class ScreenCapturerWgc : public ScreenCapturer,
+                          public WgcSession::wgc_session_observer {
+ public:
+  ScreenCapturerWgc();
+  ~ScreenCapturerWgc();
+
+ public:
+  bool IsWgcSupported();
+
+  virtual int Init(const RECORD_DESKTOP_RECT &rect, const int fps,
+                   cb_desktop_data cb);
+  virtual int Destroy();
+
+  virtual int Start();
+
+  int Pause();
+  int Resume();
+  int Stop();
+
+  void OnFrame(const WgcSession::wgc_session_frame &frame);
+
+ protected:
+  void CleanUp();
+
+ private:
+  WgcSession *session_ = nullptr;
+
+  std::atomic_bool _running;
+  std::atomic_bool _paused;
+  std::atomic_bool _inited;
+
+  std::thread _thread;
+
+  std::string _device_name;
+
+  RECORD_DESKTOP_RECT _rect;
+
+  int _fps;
+
+  cb_desktop_data _on_data;
+
+  unsigned char *nv12_frame_ = nullptr;
+};
+
+#endif

src/screen_capturer/windows/wgc_session.h

@@ -0,0 +1,40 @@
+#ifndef _WGC_SESSION_H_
+#define _WGC_SESSION_H_
+
+#include <Windows.h>
+
+class WgcSession {
+ public:
+  struct wgc_session_frame {
+    unsigned int width;
+    unsigned int height;
+    unsigned int row_pitch;
+
+    const unsigned char *data;
+  };
+
+  class wgc_session_observer {
+   public:
+    virtual ~wgc_session_observer() {}
+    virtual void OnFrame(const wgc_session_frame &frame) = 0;
+  };
+
+ public:
+  virtual void Release() = 0;
+
+  virtual int Initialize(HWND hwnd) = 0;
+  virtual int Initialize(HMONITOR hmonitor) = 0;
+
+  virtual void RegisterObserver(wgc_session_observer *observer) = 0;
+
+  virtual int Start() = 0;
+  virtual int Stop() = 0;
+
+  virtual int Pause() = 0;
+  virtual int Resume() = 0;
+
+ protected:
+  virtual ~WgcSession(){};
+};
+
+#endif

src/screen_capturer/windows/wgc_session_impl.cpp

@@ -0,0 +1,366 @@
+#include "wgc_session_impl.h"
+
+#include <Windows.Graphics.Capture.Interop.h>
+
+#include <atomic>
+#include <functional>
+#include <iostream>
+#include <memory>
+
+#define CHECK_INIT                            \
+  if (!is_initialized_) {                     \
+    std::cout << "AE_NEED_INIT" << std::endl; \
+    return 4;                                 \
+  }
+
+#define CHECK_CLOSED                         \
+  if (cleaned_.load() == true) {             \
+    throw winrt::hresult_error(RO_E_CLOSED); \
+  }
+
+extern "C" {
+HRESULT __stdcall CreateDirect3D11DeviceFromDXGIDevice(
+    ::IDXGIDevice *dxgiDevice, ::IInspectable **graphicsDevice);
+}
+
+WgcSessionImpl::WgcSessionImpl() {}
+
+WgcSessionImpl::~WgcSessionImpl() {
+  Stop();
+  CleanUp();
+}
+
+void WgcSessionImpl::Release() { delete this; }
+
+int WgcSessionImpl::Initialize(HWND hwnd) {
+  std::lock_guard locker(lock_);
+
+  target_.hwnd = hwnd;
+  target_.is_window = true;
+  return Initialize();
+}
+
+int WgcSessionImpl::Initialize(HMONITOR hmonitor) {
+  std::lock_guard locker(lock_);
+
+  target_.hmonitor = hmonitor;
+  target_.is_window = false;
+  return Initialize();
+}
+
+void WgcSessionImpl::RegisterObserver(wgc_session_observer *observer) {
+  std::lock_guard locker(lock_);
+  observer_ = observer;
+}
+
+int WgcSessionImpl::Start() {
+  std::lock_guard locker(lock_);
+
+  if (is_running_) return 0;
+
+  int error = 1;
+
+  CHECK_INIT;
+  try {
+    if (!capture_session_) {
+      auto current_size = capture_item_.Size();
+      capture_framepool_ =
+          winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool::
+              CreateFreeThreaded(d3d11_direct_device_,
+                                 winrt::Windows::Graphics::DirectX::
+                                     DirectXPixelFormat::B8G8R8A8UIntNormalized,
+                                 2, current_size);
+      capture_session_ = capture_framepool_.CreateCaptureSession(capture_item_);
+      capture_frame_size_ = current_size;
+      capture_framepool_trigger_ = capture_framepool_.FrameArrived(
+          winrt::auto_revoke, {this, &WgcSessionImpl::OnFrame});
+      capture_close_trigger_ = capture_item_.Closed(
+          winrt::auto_revoke, {this, &WgcSessionImpl::OnClosed});
+    }
+
+    if (!capture_framepool_) throw std::exception();
+
+    is_running_ = true;
+
+    // we do not need to crate a thread to enter a message loop coz we use
+    // CreateFreeThreaded instead of Create to create a capture frame pool,
+    // we need to test the performance later
+    // loop_ = std::thread(std::bind(&WgcSessionImpl::message_func, this));
+
+    capture_session_.StartCapture();
+
+    error = 0;
+  } catch (winrt::hresult_error) {
+    std::cout << "AE_WGC_CREATE_CAPTURER_FAILED" << std::endl;
+    return 86;
+  } catch (...) {
+    return 86;
+  }
+
+  return error;
+}
+
+int WgcSessionImpl::Stop() {
+  std::lock_guard locker(lock_);
+
+  CHECK_INIT;
+
+  is_running_ = false;
+
+  // if (loop_.joinable()) loop_.join();
+
+  if (capture_framepool_trigger_) capture_framepool_trigger_.revoke();
+
+  if (capture_session_) {
+    capture_session_.Close();
+    capture_session_ = nullptr;
+  }
+
+  return 0;
+}
+
+int WgcSessionImpl::Pause() {
+  std::lock_guard locker(lock_);
+
+  CHECK_INIT;
+  return 0;
+}
+
+int WgcSessionImpl::Resume() {
+  std::lock_guard locker(lock_);
+
+  CHECK_INIT;
+  return 0;
+}
+
+auto WgcSessionImpl::CreateD3D11Device() {
+  winrt::com_ptr<ID3D11Device> d3d_device;
+  HRESULT hr;
+
+  WINRT_ASSERT(!d3d_device);
+
+  D3D_DRIVER_TYPE type = D3D_DRIVER_TYPE_HARDWARE;
+  UINT flags = D3D11_CREATE_DEVICE_BGRA_SUPPORT;
+  hr = D3D11CreateDevice(nullptr, type, nullptr, flags, nullptr, 0,
+                         D3D11_SDK_VERSION, d3d_device.put(), nullptr, nullptr);
+
+  if (DXGI_ERROR_UNSUPPORTED == hr) {
+    // change D3D_DRIVER_TYPE
+    D3D_DRIVER_TYPE type = D3D_DRIVER_TYPE_WARP;
+    hr = D3D11CreateDevice(nullptr, type, nullptr, flags, nullptr, 0,
+                           D3D11_SDK_VERSION, d3d_device.put(), nullptr,
+                           nullptr);
+  }
+
+  winrt::check_hresult(hr);
+
+  winrt::com_ptr<::IInspectable> d3d11_device;
+  winrt::check_hresult(CreateDirect3D11DeviceFromDXGIDevice(
+      d3d_device.as<IDXGIDevice>().get(), d3d11_device.put()));
+  return d3d11_device
+      .as<winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice>();
+}
+
+auto WgcSessionImpl::CreateCaptureItemForWindow(HWND hwnd) {
+  auto activation_factory = winrt::get_activation_factory<
+      winrt::Windows::Graphics::Capture::GraphicsCaptureItem>();
+  auto interop_factory = activation_factory.as<IGraphicsCaptureItemInterop>();
+  winrt::Windows::Graphics::Capture::GraphicsCaptureItem item = {nullptr};
+  interop_factory->CreateForWindow(
+      hwnd,
+      winrt::guid_of<ABI::Windows::Graphics::Capture::IGraphicsCaptureItem>(),
+      reinterpret_cast<void **>(winrt::put_abi(item)));
+  return item;
+}
+
+auto WgcSessionImpl::CreateCaptureItemForMonitor(HMONITOR hmonitor) {
+  auto activation_factory = winrt::get_activation_factory<
+      winrt::Windows::Graphics::Capture::GraphicsCaptureItem>();
+  auto interop_factory = activation_factory.as<IGraphicsCaptureItemInterop>();
+  winrt::Windows::Graphics::Capture::GraphicsCaptureItem item = {nullptr};
+  interop_factory->CreateForMonitor(
+      hmonitor,
+      winrt::guid_of<ABI::Windows::Graphics::Capture::IGraphicsCaptureItem>(),
+      reinterpret_cast<void **>(winrt::put_abi(item)));
+  return item;
+}
+
+HRESULT WgcSessionImpl::CreateMappedTexture(
+    winrt::com_ptr<ID3D11Texture2D> src_texture, unsigned int width,
+    unsigned int height) {
+  D3D11_TEXTURE2D_DESC src_desc;
+  src_texture->GetDesc(&src_desc);
+  D3D11_TEXTURE2D_DESC map_desc;
+  map_desc.Width = width == 0 ? src_desc.Width : width;
+  map_desc.Height = height == 0 ? src_desc.Height : height;
+  map_desc.MipLevels = src_desc.MipLevels;
+  map_desc.ArraySize = src_desc.ArraySize;
+  map_desc.Format = src_desc.Format;
+  map_desc.SampleDesc = src_desc.SampleDesc;
+  map_desc.Usage = D3D11_USAGE_STAGING;
+  map_desc.BindFlags = 0;
+  map_desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ;
+  map_desc.MiscFlags = 0;
+
+  auto d3dDevice =
+      GetDXGIInterfaceFromObject<ID3D11Device>(d3d11_direct_device_);
+
+  return d3dDevice->CreateTexture2D(&map_desc, nullptr,
+                                    d3d11_texture_mapped_.put());
+}
+
+void WgcSessionImpl::OnFrame(
+    winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool const &sender,
+    winrt::Windows::Foundation::IInspectable const &args) {
+  std::lock_guard locker(lock_);
+
+  auto is_new_size = false;
+
+  {
+    auto frame = sender.TryGetNextFrame();
+    auto frame_size = frame.ContentSize();
+
+    if (frame_size.Width != capture_frame_size_.Width ||
+        frame_size.Height != capture_frame_size_.Height) {
+      // The thing we have been capturing has changed size.
+      // We need to resize our swap chain first, then blit the pixels.
+      // After we do that, retire the frame and then recreate our frame pool.
+      is_new_size = true;
+      capture_frame_size_ = frame_size;
+    }
+
+    // copy to mapped texture
+    {
+      auto frame_captured =
+          GetDXGIInterfaceFromObject<ID3D11Texture2D>(frame.Surface());
+
+      if (!d3d11_texture_mapped_ || is_new_size)
+        CreateMappedTexture(frame_captured);
+
+      d3d11_device_context_->CopyResource(d3d11_texture_mapped_.get(),
+                                          frame_captured.get());
+
+      D3D11_MAPPED_SUBRESOURCE map_result;
+      HRESULT hr = d3d11_device_context_->Map(
+          d3d11_texture_mapped_.get(), 0, D3D11_MAP_READ,
+          0 /*coz we use CreateFreeThreaded, so we cant use flags
+               D3D11_MAP_FLAG_DO_NOT_WAIT*/
+          ,
+          &map_result);
+      if (FAILED(hr)) {
+        OutputDebugStringW(
+            (L"map resource failed: " + std::to_wstring(hr)).c_str());
+      }
+
+      // copy data from map_result.pData
+      if (map_result.pData && observer_) {
+        observer_->OnFrame(wgc_session_frame{
+            static_cast<unsigned int>(frame_size.Width),
+            static_cast<unsigned int>(frame_size.Height), map_result.RowPitch,
+            const_cast<const unsigned char *>(
+                (unsigned char *)map_result.pData)});
+      }
+
+      d3d11_device_context_->Unmap(d3d11_texture_mapped_.get(), 0);
+    }
+  }
+
+  if (is_new_size) {
+    capture_framepool_.Recreate(d3d11_direct_device_,
+                                winrt::Windows::Graphics::DirectX::
+                                    DirectXPixelFormat::B8G8R8A8UIntNormalized,
+                                2, capture_frame_size_);
+  }
+}
+
+void WgcSessionImpl::OnClosed(
+    winrt::Windows::Graphics::Capture::GraphicsCaptureItem const &,
+    winrt::Windows::Foundation::IInspectable const &) {
+  OutputDebugStringW(L"WgcSessionImpl::OnClosed");
+}
+
+int WgcSessionImpl::Initialize() {
+  if (is_initialized_) return 0;
+
+  if (!(d3d11_direct_device_ = CreateD3D11Device())) {
+    std::cout << "AE_D3D_CREATE_DEVICE_FAILED" << std::endl;
+    return 1;
+  }
+
+  try {
+    if (target_.is_window)
+      capture_item_ = CreateCaptureItemForWindow(target_.hwnd);
+    else
+      capture_item_ = CreateCaptureItemForMonitor(target_.hmonitor);
+
+    // Set up
+    auto d3d11_device =
+        GetDXGIInterfaceFromObject<ID3D11Device>(d3d11_direct_device_);
+    d3d11_device->GetImmediateContext(d3d11_device_context_.put());
+
+  } catch (winrt::hresult_error) {
+    std::cout << "AE_WGC_CREATE_CAPTURER_FAILED" << std::endl;
+    return 86;
+  } catch (...) {
+    return 86;
+  }
+
+  is_initialized_ = true;
+
+  return 0;
+}
+
+void WgcSessionImpl::CleanUp() {
+  std::lock_guard locker(lock_);
+
+  auto expected = false;
+  if (cleaned_.compare_exchange_strong(expected, true)) {
+    capture_close_trigger_.revoke();
+    capture_framepool_trigger_.revoke();
+
+    if (capture_framepool_) capture_framepool_.Close();
+
+    if (capture_session_) capture_session_.Close();
+
+    capture_framepool_ = nullptr;
+    capture_session_ = nullptr;
+    capture_item_ = nullptr;
+
+    is_initialized_ = false;
+  }
+}
+
+LRESULT CALLBACK WindowProc(HWND window, UINT message, WPARAM w_param,
+                            LPARAM l_param) {
+  return DefWindowProc(window, message, w_param, l_param);
+}
+
+// void WgcSessionImpl::message_func() {
+//   const std::wstring kClassName = L"am_fake_window";
+
+//   WNDCLASS wc = {};
+
+//   wc.style = CS_HREDRAW | CS_VREDRAW;
+//   wc.lpfnWndProc = DefWindowProc;
+//   wc.hCursor = LoadCursor(nullptr, IDC_ARROW);
+//   wc.hbrBackground = reinterpret_cast<HBRUSH>(COLOR_WINDOW);
+//   wc.lpszClassName = kClassName.c_str();
+
+//   if (!::RegisterClassW(&wc)) return;
+
+//   hwnd_ = ::CreateWindowW(kClassName.c_str(), nullptr, WS_OVERLAPPEDWINDOW,
+//   0,
+//                           0, 0, 0, nullptr, nullptr, nullptr, nullptr);
+//   MSG msg;
+//   while (is_running_) {
+//     while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
+//       if (!is_running_) break;
+//       TranslateMessage(&msg);
+//       DispatchMessage(&msg);
+//     }
+//     Sleep(10);
+//   }
+
+//   ::CloseWindow(hwnd_);
+//   ::DestroyWindow(hwnd_);
+// }

src/screen_capturer/windows/wgc_session_impl.h

@@ -0,0 +1,116 @@
+#ifndef _WGC_SESSION_IMPL_H_
+#define _WGC_SESSION_IMPL_H_
+
+#include <d3d11_4.h>
+#include <winrt/Windows.Foundation.h>
+#include <winrt/Windows.Graphics.Capture.h>
+
+#include <mutex>
+#include <thread>
+
+#include "wgc_session.h"
+
+class WgcSessionImpl : public WgcSession {
+  struct __declspec(uuid("A9B3D012-3DF2-4EE3-B8D1-8695F457D3C1"))
+      IDirect3DDxgiInterfaceAccess : ::IUnknown {
+    virtual HRESULT __stdcall GetInterface(GUID const &id, void **object) = 0;
+  };
+
+  template <typename T>
+  inline auto GetDXGIInterfaceFromObject(
+      winrt::Windows::Foundation::IInspectable const &object) {
+    auto access = object.as<IDirect3DDxgiInterfaceAccess>();
+    winrt::com_ptr<T> result;
+    winrt::check_hresult(
+        access->GetInterface(winrt::guid_of<T>(), result.put_void()));
+    return result;
+  }
+
+  struct {
+    union {
+      HWND hwnd;
+      HMONITOR hmonitor;
+    };
+    bool is_window;
+  } target_{0};
+
+ public:
+  WgcSessionImpl();
+  ~WgcSessionImpl() override;
+
+ public:
+  void Release() override;
+
+  int Initialize(HWND hwnd) override;
+  int Initialize(HMONITOR hmonitor) override;
+
+  void RegisterObserver(wgc_session_observer *observer) override;
+
+  int Start() override;
+  int Stop() override;
+
+  int Pause() override;
+  int Resume() override;
+
+ private:
+  auto CreateD3D11Device();
+  auto CreateCaptureItemForWindow(HWND hwnd);
+  auto CreateCaptureItemForMonitor(HMONITOR hmonitor);
+
+  HRESULT CreateMappedTexture(winrt::com_ptr<ID3D11Texture2D> src_texture,
+                              unsigned int width = 0, unsigned int height = 0);
+  void OnFrame(
+      winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool const
+          &sender,
+      winrt::Windows::Foundation::IInspectable const &args);
+  void OnClosed(winrt::Windows::Graphics::Capture::GraphicsCaptureItem const &,
+                winrt::Windows::Foundation::IInspectable const &);
+
+  int Initialize();
+  void CleanUp();
+
+  // void message_func();
+
+ private:
+  std::mutex lock_;
+  bool is_initialized_ = false;
+  bool is_running_ = false;
+  bool is_paused_ = false;
+
+  wgc_session_observer *observer_ = nullptr;
+
+  // wgc
+  winrt::Windows::Graphics::Capture::GraphicsCaptureItem capture_item_{nullptr};
+  winrt::Windows::Graphics::Capture::GraphicsCaptureSession capture_session_{
+      nullptr};
+  winrt::Windows::Graphics::SizeInt32 capture_frame_size_;
+
+  winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice
+      d3d11_direct_device_{nullptr};
+  winrt::com_ptr<ID3D11DeviceContext> d3d11_device_context_{nullptr};
+  winrt::com_ptr<ID3D11Texture2D> d3d11_texture_mapped_{nullptr};
+
+  std::atomic<bool> cleaned_ = false;
+  winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool
+      capture_framepool_{nullptr};
+  winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool::
+      FrameArrived_revoker capture_framepool_trigger_;
+  winrt::Windows::Graphics::Capture::GraphicsCaptureItem::Closed_revoker
+      capture_close_trigger_;
+
+  // message loop
+  std::thread loop_;
+  HWND hwnd_ = nullptr;
+};
+
+// template <typename T>
+// inline auto WgcSessionImpl::GetDXGIInterfaceFromObject(
+//     winrt::Windows::Foundation::IInspectable const &object) {
+//   auto access = object.as<IDirect3DDxgiInterfaceAccess>();
+//   winrt::com_ptr<T> result;
+//   winrt::check_hresult(
+//       access->GetInterface(winrt::guid_of<T>(), result.put_void()));
+//   return result;
+// }
+
+#endif