Selective merge of Windows system cursor fix #66

This commit is contained in:
webadderall
2026-03-19 11:10:44 +11:00
parent ab0e5880a4
commit 6c06c96f91
12 changed files with 1259 additions and 5 deletions
+3 -3
View File
@@ -848,7 +848,7 @@ async function buildFfmpegCaptureArgs(source: SelectedSource, outputPath: string
}
function getWindowsCaptureExePath() {
return resolveUnpackedAppPath('electron', 'native', 'windows-capture', 'build', 'Release', 'windows-capture.exe')
return resolveUnpackedAppPath('electron', 'native', 'wgc-capture', 'build', 'Release', 'wgc-capture.exe')
}
function getCursorMonitorExePath() {
@@ -865,8 +865,8 @@ async function isNativeWindowsCaptureAvailable(): Promise<boolean> {
}
const os = await import('node:os')
const [major, minor] = os.release().split('.').map(Number)
return major > 6 || (major === 6 && minor >= 2)
const [major, , build] = os.release().split('.').map(Number)
return major >= 10 && build >= 19041
}
function waitForWindowsCaptureStart(proc: ChildProcessWithoutNullStreams) {
@@ -0,0 +1,27 @@
cmake_minimum_required(VERSION 3.20)
project(wgc-capture LANGUAGES CXX)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
add_executable(wgc-capture
src/main.cpp
src/wgc_session.cpp
src/mf_encoder.cpp
src/monitor_utils.cpp
src/wasapi_loopback.cpp
)
target_compile_options(wgc-capture PRIVATE /EHsc /W3 /utf-8)
target_link_libraries(wgc-capture PRIVATE
windowsapp
d3d11
dxgi
mfplat
mfreadwrite
mf
mfuuid
ole32
shcore
)
+329
View File
@@ -0,0 +1,329 @@
#include "wgc_session.h"
#include "mf_encoder.h"
#include "monitor_utils.h"
#include "wasapi_loopback.h"
#include <winrt/Windows.Foundation.h>
#include <winrt/Windows.System.h>
#include <iostream>
#include <string>
#include <thread>
#include <atomic>
#include <mutex>
#include <condition_variable>
#include <chrono>
static std::atomic<bool> g_stopRequested{false};
static std::atomic<bool> g_pauseRequested{false};
static std::atomic<bool> g_resumePending{false};
static std::atomic<int64_t> g_lastFrameTimestampHns{0};
static std::atomic<int64_t> g_pauseStartTimestampHns{0};
static std::atomic<int64_t> g_accumulatedPausedHns{0};
static std::mutex g_stopMutex;
static std::condition_variable g_stopCv;
struct CaptureConfig {
int displayId = 0;
int64_t windowHandle = 0;
std::string outputPath;
std::string audioOutputPath;
std::string micOutputPath;
std::string micDeviceName;
int fps = 60;
int width = 0;
int height = 0;
bool captureSystemAudio = false;
bool captureMic = false;
};
static bool parseSimpleJson(const std::string& json, CaptureConfig& config) {
auto findInt = [&](const std::string& key) -> int {
auto pos = json.find("\"" + key + "\"");
if (pos == std::string::npos) return -1;
pos = json.find(':', pos);
if (pos == std::string::npos) return -1;
pos++;
while (pos < json.size() && (json[pos] == ' ' || json[pos] == '\t')) pos++;
try {
return std::stoi(json.substr(pos));
} catch (...) {
return -1;
}
};
auto findInt64 = [&](const std::string& key) -> int64_t {
auto pos = json.find("\"" + key + "\"");
if (pos == std::string::npos) return -1;
pos = json.find(':', pos);
if (pos == std::string::npos) return -1;
pos++;
while (pos < json.size() && (json[pos] == ' ' || json[pos] == '\t')) pos++;
try {
return std::stoll(json.substr(pos));
} catch (...) {
return -1;
}
};
auto findString = [&](const std::string& key) -> std::string {
auto pos = json.find("\"" + key + "\"");
if (pos == std::string::npos) return "";
pos = json.find(':', pos);
if (pos == std::string::npos) return "";
pos++;
while (pos < json.size() && (json[pos] == ' ' || json[pos] == '\t')) pos++;
if (pos >= json.size() || json[pos] != '"') return "";
pos++;
std::string result;
while (pos < json.size() && json[pos] != '"') {
if (json[pos] == '\\' && pos + 1 < json.size()) {
pos++;
if (json[pos] == 'n') result += '\n';
else if (json[pos] == 't') result += '\t';
else if (json[pos] == '\\') result += '\\';
else if (json[pos] == '"') result += '"';
else if (json[pos] == '/') result += '/';
else result += json[pos];
} else {
result += json[pos];
}
pos++;
}
return result;
};
config.outputPath = findString("outputPath");
if (config.outputPath.empty()) return false;
int displayId = findInt("displayId");
if (displayId >= 0) config.displayId = displayId;
int64_t windowHandle = findInt64("windowHandle");
if (windowHandle > 0) config.windowHandle = windowHandle;
int fps = findInt("fps");
if (fps > 0) config.fps = fps;
int width = findInt("width");
if (width > 0) config.width = width;
int height = findInt("height");
if (height > 0) config.height = height;
config.audioOutputPath = findString("audioOutputPath");
config.micOutputPath = findString("micOutputPath");
config.micDeviceName = findString("micDeviceName");
auto findBool = [&](const std::string& key) -> bool {
auto pos = json.find("\"" + key + "\"");
if (pos == std::string::npos) return false;
auto colonPos = json.find(':', pos);
if (colonPos == std::string::npos) return false;
auto valStart = json.find_first_not_of(" \t", colonPos + 1);
return valStart != std::string::npos && json.substr(valStart, 4) == "true";
};
config.captureSystemAudio = findBool("captureSystemAudio");
config.captureMic = findBool("captureMic");
return true;
}
static std::wstring utf8ToWide(const std::string& str) {
if (str.empty()) return L"";
int len = MultiByteToWideChar(CP_UTF8, 0, str.c_str(), static_cast<int>(str.size()), nullptr, 0);
std::wstring wstr(len, L'\0');
MultiByteToWideChar(CP_UTF8, 0, str.c_str(), static_cast<int>(str.size()), &wstr[0], len);
return wstr;
}
static void stdinListenerThread() {
std::string line;
while (std::getline(std::cin, line)) {
// Trim whitespace
while (!line.empty() && (line.back() == '\r' || line.back() == '\n' || line.back() == ' ')) {
line.pop_back();
}
if (line == "pause") {
g_pauseRequested = true;
g_pauseStartTimestampHns = g_lastFrameTimestampHns.load();
continue;
}
if (line == "resume") {
g_pauseRequested = false;
g_resumePending = true;
continue;
}
if (line == "stop") {
g_stopRequested = true;
g_stopCv.notify_all();
return;
}
}
// stdin closed (parent process died)
g_stopRequested = true;
g_stopCv.notify_all();
}
int main(int argc, char* argv[]) {
if (argc < 2) {
std::cerr << "ERROR: Missing JSON config argument" << std::endl;
return 1;
}
winrt::init_apartment(winrt::apartment_type::multi_threaded);
CaptureConfig config;
if (!parseSimpleJson(argv[1], config)) {
std::cerr << "ERROR: Failed to parse config JSON" << std::endl;
return 1;
}
WgcSession session;
if (config.windowHandle > 0) {
HWND hwnd = reinterpret_cast<HWND>(static_cast<intptr_t>(config.windowHandle));
if (!IsWindow(hwnd)) {
std::cerr << "ERROR: Invalid window handle " << config.windowHandle << std::endl;
return 1;
}
if (!session.initialize(hwnd, config.fps)) {
std::cerr << "ERROR: Failed to initialize WGC window capture session" << std::endl;
return 1;
}
} else {
HMONITOR monitor = findMonitorByDisplayId(config.displayId);
if (!monitor) {
std::cerr << "ERROR: Could not find monitor for displayId " << config.displayId << std::endl;
return 1;
}
if (!session.initialize(monitor, config.fps)) {
std::cerr << "ERROR: Failed to initialize WGC capture session" << std::endl;
return 1;
}
}
int captureWidth = config.width > 0 ? config.width : session.captureWidth();
int captureHeight = config.height > 0 ? config.height : session.captureHeight();
// Ensure even dimensions for H.264
captureWidth = (captureWidth / 2) * 2;
captureHeight = (captureHeight / 2) * 2;
// Initialize encoder
MFEncoder encoder;
std::wstring outputPathW = utf8ToWide(config.outputPath);
if (!encoder.initialize(outputPathW, captureWidth, captureHeight, config.fps,
session.device(), session.context())) {
std::cerr << "ERROR: Failed to initialize Media Foundation encoder" << std::endl;
return 1;
}
// Set up frame callback
std::atomic<int64_t> frameCount{0};
session.setFrameCallback([&](ID3D11Texture2D* texture, int64_t timestampHns) {
g_lastFrameTimestampHns = timestampHns;
if (g_stopRequested) return;
if (g_pauseRequested) return;
int64_t adjustedTimestampHns = timestampHns;
if (g_resumePending.exchange(false)) {
const int64_t pauseStart = g_pauseStartTimestampHns.load();
if (pauseStart > 0 && timestampHns > pauseStart) {
g_accumulatedPausedHns += (timestampHns - pauseStart);
}
}
adjustedTimestampHns -= g_accumulatedPausedHns.load();
if (adjustedTimestampHns < 0) {
adjustedTimestampHns = 0;
}
if (encoder.writeFrame(texture, adjustedTimestampHns)) {
frameCount++;
}
});
// Start stdin listener
std::thread stdinThread(stdinListenerThread);
stdinThread.detach();
// Initialize WASAPI captures (but don't start yet)
WasapiCapture loopback;
WasapiCapture micCapture;
bool audioActive = false;
bool audioInitialized = false;
bool micActive = false;
bool micInitialized = false;
if (config.captureSystemAudio && !config.audioOutputPath.empty()) {
audioInitialized = loopback.initializeLoopback(config.audioOutputPath);
if (!audioInitialized) {
std::cerr << "WARNING: Failed to initialize WASAPI loopback" << std::endl;
}
}
if (config.captureMic && !config.micOutputPath.empty()) {
micInitialized = micCapture.initializeMic(config.micOutputPath, config.micDeviceName);
if (!micInitialized) {
std::cerr << "WARNING: Failed to initialize WASAPI mic capture" << std::endl;
}
}
// Start video capture, then audio immediately after for sync
if (!session.startCapture()) {
std::cerr << "ERROR: Failed to start WGC capture" << std::endl;
return 1;
}
if (audioInitialized) {
audioActive = loopback.start();
}
if (micInitialized) {
micActive = micCapture.start();
}
std::cout << "Recording started" << std::endl;
std::cout.flush();
// Wait for stop signal while pausing/resuming audio tracks in lockstep.
while (!g_stopRequested) {
if (g_pauseRequested) {
if (audioActive) loopback.pause();
if (micActive) micCapture.pause();
} else {
if (audioActive) loopback.resume();
if (micActive) micCapture.resume();
}
std::unique_lock<std::mutex> lock(g_stopMutex);
g_stopCv.wait_for(lock, std::chrono::milliseconds(20), [] { return g_stopRequested.load(); });
}
// Stop capture and finalize
session.stopCapture();
if (audioActive) loopback.stop();
if (micActive) micCapture.stop();
encoder.finalize();
std::cout << "Recording stopped. Output path: " << config.outputPath << std::endl;
if (audioActive) {
std::cout << "Audio path: " << config.audioOutputPath << std::endl;
}
if (micActive) {
std::cout << "Mic path: " << config.micOutputPath << std::endl;
}
std::cout.flush();
// Allow pipe buffers to drain before forceful exit
Sleep(100);
// Fast exit to avoid WinRT/COM teardown crashes during apartment cleanup
ExitProcess(0);
}
@@ -0,0 +1,205 @@
#include "mf_encoder.h"
#include <mfapi.h>
#include <mferror.h>
#include <codecapi.h>
#include <iostream>
#include <cstring>
#pragma comment(lib, "mfplat.lib")
#pragma comment(lib, "mfreadwrite.lib")
#pragma comment(lib, "mf.lib")
#pragma comment(lib, "mfuuid.lib")
static int clampByte(int v) {
return v < 0 ? 0 : (v > 255 ? 255 : v);
}
MFEncoder::MFEncoder() {}
MFEncoder::~MFEncoder() {
finalize();
}
bool MFEncoder::initialize(const std::wstring& outputPath, int width, int height, int fps,
ID3D11Device* device, ID3D11DeviceContext* context) {
if (initialized_) return false;
if (width % 2 != 0 || height % 2 != 0) {
std::cerr << "ERROR: Encoder dimensions must be even, got " << width << "x" << height << std::endl;
return false;
}
width_ = width;
height_ = height;
fps_ = fps;
device_ = device;
context_ = context;
HRESULT hr = MFStartup(MF_VERSION);
if (FAILED(hr)) {
std::cerr << "ERROR: MFStartup failed: 0x" << std::hex << hr << std::endl;
return false;
}
// Output media type (H.264)
ComPtr<IMFMediaType> outputType;
hr = MFCreateMediaType(&outputType);
if (FAILED(hr)) return false;
outputType->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
outputType->SetGUID(MF_MT_SUBTYPE, MFVideoFormat_H264);
outputType->SetUINT32(MF_MT_AVG_BITRATE, 20000000);
MFSetAttributeSize(outputType.Get(), MF_MT_FRAME_SIZE, width_, height_);
MFSetAttributeRatio(outputType.Get(), MF_MT_FRAME_RATE, fps_, 1);
MFSetAttributeRatio(outputType.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
outputType->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
// Input media type (NV12)
ComPtr<IMFMediaType> inputType;
hr = MFCreateMediaType(&inputType);
if (FAILED(hr)) return false;
inputType->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
inputType->SetGUID(MF_MT_SUBTYPE, MFVideoFormat_NV12);
MFSetAttributeSize(inputType.Get(), MF_MT_FRAME_SIZE, width_, height_);
MFSetAttributeRatio(inputType.Get(), MF_MT_FRAME_RATE, fps_, 1);
MFSetAttributeRatio(inputType.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
inputType->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
// Create SinkWriter with MPEG4 container
ComPtr<IMFAttributes> writerAttrs;
hr = MFCreateAttributes(&writerAttrs, 1);
if (FAILED(hr)) return false;
writerAttrs->SetUINT32(MF_READWRITE_ENABLE_HARDWARE_TRANSFORMS, TRUE);
hr = MFCreateSinkWriterFromURL(outputPath.c_str(), nullptr, writerAttrs.Get(), &sinkWriter_);
if (FAILED(hr)) {
std::cerr << "ERROR: MFCreateSinkWriterFromURL failed: 0x" << std::hex << hr << std::endl;
return false;
}
hr = sinkWriter_->AddStream(outputType.Get(), &streamIndex_);
if (FAILED(hr)) {
std::cerr << "ERROR: AddStream failed: 0x" << std::hex << hr << std::endl;
return false;
}
hr = sinkWriter_->SetInputMediaType(streamIndex_, inputType.Get(), nullptr);
if (FAILED(hr)) {
std::cerr << "ERROR: SetInputMediaType failed: 0x" << std::hex << hr << std::endl;
return false;
}
hr = sinkWriter_->BeginWriting();
if (FAILED(hr)) {
std::cerr << "ERROR: BeginWriting failed: 0x" << std::hex << hr << std::endl;
return false;
}
// Pre-allocate staging texture
D3D11_TEXTURE2D_DESC stagingDesc = {};
stagingDesc.Width = width_;
stagingDesc.Height = height_;
stagingDesc.MipLevels = 1;
stagingDesc.ArraySize = 1;
stagingDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
stagingDesc.SampleDesc.Count = 1;
stagingDesc.Usage = D3D11_USAGE_STAGING;
stagingDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ;
hr = device_->CreateTexture2D(&stagingDesc, nullptr, &stagingTexture_);
if (FAILED(hr)) {
std::cerr << "ERROR: Failed to create staging texture: 0x" << std::hex << hr << std::endl;
return false;
}
// Pre-allocate NV12 buffer
const int ySize = width_ * height_;
const int uvSize = (width_ / 2) * (height_ / 2) * 2;
nv12Buffer_.resize(ySize + uvSize);
initialized_ = true;
return true;
}
bool MFEncoder::writeFrame(ID3D11Texture2D* texture, int64_t timestampHns) {
if (!initialized_ || !sinkWriter_) return false;
context_->CopyResource(stagingTexture_.Get(), texture);
D3D11_MAPPED_SUBRESOURCE mapped;
HRESULT hr = context_->Map(stagingTexture_.Get(), 0, D3D11_MAP_READ, 0, &mapped);
if (FAILED(hr)) return false;
// Convert BGRA → NV12
const uint8_t* bgra = static_cast<const uint8_t*>(mapped.pData);
const int bgraPitch = static_cast<int>(mapped.RowPitch);
// Y plane
for (int y = 0; y < height_; y++) {
for (int x = 0; x < width_; x++) {
const uint8_t* pixel = bgra + y * bgraPitch + x * 4;
uint8_t b = pixel[0], g = pixel[1], r = pixel[2];
int yVal = ((66 * r + 129 * g + 25 * b + 128) >> 8) + 16;
nv12Buffer_[y * width_ + x] = static_cast<uint8_t>(clampByte(yVal));
}
}
// UV plane (interleaved, subsampled 2x2)
const int ySize = width_ * height_;
uint8_t* uvPlane = nv12Buffer_.data() + ySize;
for (int y = 0; y < height_; y += 2) {
for (int x = 0; x < width_; x += 2) {
const uint8_t* pixel = bgra + y * bgraPitch + x * 4;
uint8_t b = pixel[0], g = pixel[1], r = pixel[2];
int u = ((-38 * r - 74 * g + 112 * b + 128) >> 8) + 128;
int v = ((112 * r - 94 * g - 18 * b + 128) >> 8) + 128;
int uvIdx = (y / 2) * width_ + (x / 2) * 2;
uvPlane[uvIdx] = static_cast<uint8_t>(clampByte(u));
uvPlane[uvIdx + 1] = static_cast<uint8_t>(clampByte(v));
}
}
context_->Unmap(stagingTexture_.Get(), 0);
// Create MF sample
DWORD bufferSize = static_cast<DWORD>(nv12Buffer_.size());
ComPtr<IMFMediaBuffer> buffer;
hr = MFCreateMemoryBuffer(bufferSize, &buffer);
if (FAILED(hr)) return false;
BYTE* bufferData = nullptr;
hr = buffer->Lock(&bufferData, nullptr, nullptr);
if (FAILED(hr)) return false;
std::memcpy(bufferData, nv12Buffer_.data(), bufferSize);
buffer->Unlock();
buffer->SetCurrentLength(bufferSize);
ComPtr<IMFSample> sample;
hr = MFCreateSample(&sample);
if (FAILED(hr)) return false;
sample->AddBuffer(buffer.Get());
sample->SetSampleTime(timestampHns);
sample->SetSampleDuration(10000000LL / fps_);
hr = sinkWriter_->WriteSample(streamIndex_, sample.Get());
return SUCCEEDED(hr);
}
bool MFEncoder::finalize() {
if (!initialized_) return false;
initialized_ = false;
stagingTexture_.Reset();
nv12Buffer_.clear();
nv12Buffer_.shrink_to_fit();
if (!sinkWriter_) return false;
HRESULT hr = sinkWriter_->Finalize();
sinkWriter_.Reset();
MFShutdown();
return SUCCEEDED(hr);
}
@@ -0,0 +1,35 @@
#pragma once
#include <windows.h>
#include <mfapi.h>
#include <mfidl.h>
#include <mfreadwrite.h>
#include <d3d11.h>
#include <wrl/client.h>
#include <string>
#include <vector>
using Microsoft::WRL::ComPtr;
class MFEncoder {
public:
MFEncoder();
~MFEncoder();
bool initialize(const std::wstring& outputPath, int width, int height, int fps,
ID3D11Device* device, ID3D11DeviceContext* context);
bool writeFrame(ID3D11Texture2D* texture, int64_t timestampHns);
bool finalize();
private:
ComPtr<IMFSinkWriter> sinkWriter_;
ID3D11Device* device_ = nullptr;
ID3D11DeviceContext* context_ = nullptr;
ComPtr<ID3D11Texture2D> stagingTexture_;
std::vector<uint8_t> nv12Buffer_;
DWORD streamIndex_ = 0;
int width_ = 0;
int height_ = 0;
int fps_ = 60;
bool initialized_ = false;
};
@@ -0,0 +1,61 @@
#include "monitor_utils.h"
#include <ShellScalingApi.h>
static BOOL CALLBACK enumMonitorCallback(HMONITOR hMonitor, HDC, LPRECT, LPARAM lParam) {
auto* monitors = reinterpret_cast<std::vector<MonitorInfo>*>(lParam);
MONITORINFOEXW mi = {};
mi.cbSize = sizeof(mi);
if (GetMonitorInfoW(hMonitor, &mi)) {
MonitorInfo info;
info.handle = hMonitor;
info.x = mi.rcMonitor.left;
info.y = mi.rcMonitor.top;
info.width = mi.rcMonitor.right - mi.rcMonitor.left;
info.height = mi.rcMonitor.bottom - mi.rcMonitor.top;
info.deviceName = mi.szDevice;
monitors->push_back(info);
}
return TRUE;
}
std::vector<MonitorInfo> enumerateMonitors() {
std::vector<MonitorInfo> monitors;
EnumDisplayMonitors(nullptr, nullptr, enumMonitorCallback, reinterpret_cast<LPARAM>(&monitors));
return monitors;
}
// Electron uses the HMONITOR handle value cast to a number as the display ID.
HMONITOR findMonitorByDisplayId(int displayId) {
auto monitors = enumerateMonitors();
for (const auto& m : monitors) {
if (static_cast<int>(reinterpret_cast<intptr_t>(m.handle)) == displayId) {
return m.handle;
}
}
if (!monitors.empty()) {
return monitors[0].handle;
}
return MonitorFromPoint({0, 0}, MONITOR_DEFAULTTOPRIMARY);
}
MonitorInfo getMonitorInfo(HMONITOR monitor) {
MonitorInfo info;
info.handle = monitor;
MONITORINFOEXW mi = {};
mi.cbSize = sizeof(mi);
if (GetMonitorInfoW(monitor, &mi)) {
info.x = mi.rcMonitor.left;
info.y = mi.rcMonitor.top;
info.width = mi.rcMonitor.right - mi.rcMonitor.left;
info.height = mi.rcMonitor.bottom - mi.rcMonitor.top;
info.deviceName = mi.szDevice;
}
return info;
}
@@ -0,0 +1,18 @@
#pragma once
#include <windows.h>
#include <string>
#include <vector>
struct MonitorInfo {
HMONITOR handle;
int x;
int y;
int width;
int height;
std::wstring deviceName;
};
std::vector<MonitorInfo> enumerateMonitors();
HMONITOR findMonitorByDisplayId(int displayId);
MonitorInfo getMonitorInfo(HMONITOR monitor);
@@ -0,0 +1,242 @@
#include "wasapi_loopback.h"
#include <functiondiscoverykeys_devpkey.h>
#include <iostream>
#include <cstring>
#pragma comment(lib, "ole32.lib")
static const CLSID CLSID_MMDeviceEnumerator_ = __uuidof(MMDeviceEnumerator);
static const IID IID_IMMDeviceEnumerator_ = __uuidof(IMMDeviceEnumerator);
static const IID IID_IAudioClient_ = __uuidof(IAudioClient);
static const IID IID_IAudioCaptureClient_ = __uuidof(IAudioCaptureClient);
WasapiCapture::WasapiCapture() {}
WasapiCapture::~WasapiCapture() {
stop();
if (mixFormat_) CoTaskMemFree(mixFormat_);
if (captureClient_) captureClient_->Release();
if (audioClient_) audioClient_->Release();
if (device_) device_->Release();
if (enumerator_) enumerator_->Release();
}
static std::wstring utf8ToWide(const std::string& str) {
if (str.empty()) return L"";
int len = MultiByteToWideChar(CP_UTF8, 0, str.c_str(), static_cast<int>(str.size()), nullptr, 0);
std::wstring wstr(len, L'\0');
MultiByteToWideChar(CP_UTF8, 0, str.c_str(), static_cast<int>(str.size()), &wstr[0], len);
return wstr;
}
IMMDevice* WasapiCapture::findCaptureDeviceByName(const std::wstring& targetName) {
IMMDeviceCollection* collection = nullptr;
HRESULT hr = enumerator_->EnumAudioEndpoints(eCapture, DEVICE_STATE_ACTIVE, &collection);
if (FAILED(hr)) return nullptr;
UINT count = 0;
collection->GetCount(&count);
for (UINT i = 0; i < count; i++) {
IMMDevice* dev = nullptr;
collection->Item(i, &dev);
IPropertyStore* store = nullptr;
dev->OpenPropertyStore(STGM_READ, &store);
PROPVARIANT pv;
PropVariantInit(&pv);
store->GetValue(PKEY_Device_FriendlyName, &pv);
std::wstring name = pv.pwszVal ? pv.pwszVal : L"";
PropVariantClear(&pv);
store->Release();
if (name.find(targetName) != std::wstring::npos || targetName.find(name) != std::wstring::npos) {
collection->Release();
return dev;
}
dev->Release();
}
collection->Release();
return nullptr;
}
bool WasapiCapture::initializeLoopback(const std::string& outputPath) {
outputPath_ = outputPath;
streamFlags_ = AUDCLNT_STREAMFLAGS_LOOPBACK;
HRESULT hr = CoCreateInstance(
CLSID_MMDeviceEnumerator_, nullptr, CLSCTX_ALL,
IID_IMMDeviceEnumerator_, reinterpret_cast<void**>(&enumerator_));
if (FAILED(hr)) return false;
hr = enumerator_->GetDefaultAudioEndpoint(eRender, eConsole, &device_);
if (FAILED(hr)) return false;
return initializeCommon();
}
bool WasapiCapture::initializeMic(const std::string& outputPath, const std::string& deviceName) {
outputPath_ = outputPath;
streamFlags_ = 0;
HRESULT hr = CoCreateInstance(
CLSID_MMDeviceEnumerator_, nullptr, CLSCTX_ALL,
IID_IMMDeviceEnumerator_, reinterpret_cast<void**>(&enumerator_));
if (FAILED(hr)) return false;
if (!deviceName.empty()) {
device_ = findCaptureDeviceByName(utf8ToWide(deviceName));
}
if (!device_) {
hr = enumerator_->GetDefaultAudioEndpoint(eCapture, eCommunications, &device_);
if (FAILED(hr)) {
hr = enumerator_->GetDefaultAudioEndpoint(eCapture, eConsole, &device_);
if (FAILED(hr)) return false;
}
}
return initializeCommon();
}
bool WasapiCapture::initializeCommon() {
HRESULT hr = device_->Activate(IID_IAudioClient_, CLSCTX_ALL, nullptr,
reinterpret_cast<void**>(&audioClient_));
if (FAILED(hr)) return false;
hr = audioClient_->GetMixFormat(&mixFormat_);
if (FAILED(hr)) return false;
REFERENCE_TIME bufferDuration = 200000; // 20ms
hr = audioClient_->Initialize(
AUDCLNT_SHAREMODE_SHARED,
streamFlags_,
bufferDuration, 0, mixFormat_, nullptr);
if (FAILED(hr)) return false;
hr = audioClient_->GetBufferSize(&bufferFrameCount_);
if (FAILED(hr)) return false;
hr = audioClient_->GetService(IID_IAudioCaptureClient_,
reinterpret_cast<void**>(&captureClient_));
if (FAILED(hr)) return false;
return true;
}
bool WasapiCapture::start() {
HRESULT hr = audioClient_->Start();
if (FAILED(hr)) return false;
capturing_ = true;
thread_ = std::thread(&WasapiCapture::captureThread, this);
return true;
}
void WasapiCapture::stop() {
if (!capturing_) return;
capturing_ = false;
if (thread_.joinable()) thread_.join();
if (audioClient_) audioClient_->Stop();
}
static int16_t floatToInt16(float v) {
v = v < -1.0f ? -1.0f : (v > 1.0f ? 1.0f : v);
return static_cast<int16_t>(v * 32767.0f);
}
bool WasapiCapture::writeWavHeader(HANDLE file, DWORD dataSize) {
WORD channels = static_cast<WORD>(mixFormat_->nChannels);
DWORD sampleRate = mixFormat_->nSamplesPerSec;
WORD bitsPerSample = 16;
WORD blockAlign = channels * (bitsPerSample / 8);
DWORD byteRate = sampleRate * blockAlign;
DWORD written;
WriteFile(file, "RIFF", 4, &written, nullptr);
DWORD chunkSize = 36 + dataSize;
WriteFile(file, &chunkSize, 4, &written, nullptr);
WriteFile(file, "WAVE", 4, &written, nullptr);
WriteFile(file, "fmt ", 4, &written, nullptr);
DWORD fmtSize = 16;
WriteFile(file, &fmtSize, 4, &written, nullptr);
WORD audioFormat = 1;
WriteFile(file, &audioFormat, 2, &written, nullptr);
WriteFile(file, &channels, 2, &written, nullptr);
WriteFile(file, &sampleRate, 4, &written, nullptr);
WriteFile(file, &byteRate, 4, &written, nullptr);
WriteFile(file, &blockAlign, 2, &written, nullptr);
WriteFile(file, &bitsPerSample, 2, &written, nullptr);
WriteFile(file, "data", 4, &written, nullptr);
WriteFile(file, &dataSize, 4, &written, nullptr);
return true;
}
void WasapiCapture::captureThread() {
WORD channels = static_cast<WORD>(mixFormat_->nChannels);
bool isFloat = (mixFormat_->wFormatTag == WAVE_FORMAT_IEEE_FLOAT) ||
(mixFormat_->wFormatTag == WAVE_FORMAT_EXTENSIBLE &&
reinterpret_cast<WAVEFORMATEXTENSIBLE*>(mixFormat_)->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT);
std::vector<int16_t> pcmBuffer;
HANDLE tmpFile = CreateFileA(
outputPath_.c_str(), GENERIC_WRITE, 0, nullptr,
CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, nullptr);
if (tmpFile == INVALID_HANDLE_VALUE) {
std::cerr << "ERROR: Cannot create audio output file" << std::endl;
return;
}
writeWavHeader(tmpFile, 0);
DWORD totalDataBytes = 0;
DWORD sleepMs = static_cast<DWORD>((static_cast<double>(bufferFrameCount_) / mixFormat_->nSamplesPerSec) * 500.0);
if (sleepMs < 5) sleepMs = 5;
while (capturing_) {
Sleep(sleepMs);
UINT32 packetLength = 0;
HRESULT hr = captureClient_->GetNextPacketSize(&packetLength);
if (FAILED(hr)) break;
while (packetLength > 0) {
BYTE* data = nullptr;
UINT32 numFrames = 0;
DWORD flags = 0;
hr = captureClient_->GetBuffer(&data, &numFrames, &flags, nullptr, nullptr);
if (FAILED(hr)) break;
UINT32 totalSamples = numFrames * channels;
if (flags & AUDCLNT_BUFFERFLAGS_SILENT) {
pcmBuffer.assign(totalSamples, 0);
} else if (isFloat) {
pcmBuffer.resize(totalSamples);
const float* src = reinterpret_cast<const float*>(data);
for (UINT32 i = 0; i < totalSamples; i++) {
pcmBuffer[i] = floatToInt16(src[i]);
}
} else {
pcmBuffer.resize(totalSamples);
std::memcpy(pcmBuffer.data(), data, totalSamples * sizeof(int16_t));
}
captureClient_->ReleaseBuffer(numFrames);
DWORD bytesToWrite = totalSamples * sizeof(int16_t);
DWORD written;
WriteFile(tmpFile, pcmBuffer.data(), bytesToWrite, &written, nullptr);
totalDataBytes += written;
hr = captureClient_->GetNextPacketSize(&packetLength);
if (FAILED(hr)) break;
}
}
SetFilePointer(tmpFile, 0, nullptr, FILE_BEGIN);
writeWavHeader(tmpFile, totalDataBytes);
CloseHandle(tmpFile);
}
@@ -0,0 +1,39 @@
#pragma once
#include <windows.h>
#include <mmdeviceapi.h>
#include <audioclient.h>
#include <string>
#include <thread>
#include <atomic>
#include <vector>
class WasapiCapture {
public:
WasapiCapture();
~WasapiCapture();
bool initializeLoopback(const std::string& outputPath);
bool initializeMic(const std::string& outputPath, const std::string& deviceName = "");
bool start();
void stop();
private:
bool initializeCommon();
void captureThread();
bool writeWavHeader(HANDLE file, DWORD dataSize);
IMMDevice* findCaptureDeviceByName(const std::wstring& name);
std::string outputPath_;
std::thread thread_;
std::atomic<bool> capturing_{false};
IMMDeviceEnumerator* enumerator_ = nullptr;
IMMDevice* device_ = nullptr;
IAudioClient* audioClient_ = nullptr;
IAudioCaptureClient* captureClient_ = nullptr;
WAVEFORMATEX* mixFormat_ = nullptr;
DWORD streamFlags_ = 0;
UINT32 bufferFrameCount_ = 0;
};
@@ -0,0 +1,236 @@
#include "wgc_session.h"
#include <windows.graphics.capture.interop.h>
#include <Windows.Graphics.Capture.h>
#include <inspectable.h>
#include <winrt/Windows.Foundation.h>
#include <winrt/Windows.System.h>
#include <iostream>
#include <chrono>
// IDirect3DDxgiInterfaceAccess is a COM interface for getting the DXGI interface
// from a WinRT IDirect3DSurface
MIDL_INTERFACE("A9B3D012-3DF2-4EE3-B8D1-8695F457D3C1")
IDirect3DDxgiInterfaceAccess : public IUnknown {
virtual HRESULT STDMETHODCALLTYPE GetInterface(REFIID iid, void** p) = 0;
};
// Convert ID3D11Device → IDirect3DDevice (WinRT interop)
extern "C" {
HRESULT __stdcall CreateDirect3D11DeviceFromDXGIDevice(
IDXGIDevice* dxgiDevice,
IInspectable** graphicsDevice);
}
WgcSession::WgcSession() {}
WgcSession::~WgcSession() {
stopCapture();
}
bool WgcSession::createD3DDevice() {
D3D_FEATURE_LEVEL featureLevels[] = {
D3D_FEATURE_LEVEL_11_1,
D3D_FEATURE_LEVEL_11_0,
};
HRESULT hr = D3D11CreateDevice(
nullptr,
D3D_DRIVER_TYPE_HARDWARE,
nullptr,
D3D11_CREATE_DEVICE_BGRA_SUPPORT,
featureLevels,
ARRAYSIZE(featureLevels),
D3D11_SDK_VERSION,
&d3dDevice_,
nullptr,
&d3dContext_);
if (FAILED(hr)) {
std::cerr << "ERROR: D3D11CreateDevice failed: 0x" << std::hex << hr << std::endl;
return false;
}
return true;
}
winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice WgcSession::createWinRTDevice() {
ComPtr<IDXGIDevice> dxgiDevice;
HRESULT hr = d3dDevice_.As(&dxgiDevice);
if (FAILED(hr)) return nullptr;
winrt::com_ptr<IInspectable> inspectable;
hr = CreateDirect3D11DeviceFromDXGIDevice(dxgiDevice.Get(), inspectable.put());
if (FAILED(hr)) return nullptr;
return inspectable.as<winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice>();
}
winrt::Windows::Graphics::Capture::GraphicsCaptureItem WgcSession::createCaptureItemForMonitor(HMONITOR monitor) {
auto factory = winrt::get_activation_factory<
winrt::Windows::Graphics::Capture::GraphicsCaptureItem>();
auto interop = factory.as<IGraphicsCaptureItemInterop>();
winrt::Windows::Graphics::Capture::GraphicsCaptureItem item{nullptr};
HRESULT hr = interop->CreateForMonitor(
monitor,
winrt::guid_of<ABI::Windows::Graphics::Capture::IGraphicsCaptureItem>(),
winrt::put_abi(item));
if (FAILED(hr)) {
std::cerr << "ERROR: CreateForMonitor failed: 0x" << std::hex << hr << std::endl;
return nullptr;
}
return item;
}
winrt::Windows::Graphics::Capture::GraphicsCaptureItem WgcSession::createCaptureItemForWindow(HWND hwnd) {
auto factory = winrt::get_activation_factory<
winrt::Windows::Graphics::Capture::GraphicsCaptureItem>();
auto interop = factory.as<IGraphicsCaptureItemInterop>();
winrt::Windows::Graphics::Capture::GraphicsCaptureItem item{nullptr};
HRESULT hr = interop->CreateForWindow(
hwnd,
winrt::guid_of<ABI::Windows::Graphics::Capture::IGraphicsCaptureItem>(),
winrt::put_abi(item));
if (FAILED(hr)) {
std::cerr << "ERROR: CreateForWindow failed: 0x" << std::hex << hr << std::endl;
return nullptr;
}
return item;
}
bool WgcSession::initializeWithItem(int fps) {
if (!captureItem_) return false;
auto size = captureItem_.Size();
captureWidth_ = size.Width;
captureHeight_ = size.Height;
framePool_ = winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool::CreateFreeThreaded(
winrtDevice_,
winrt::Windows::Graphics::DirectX::DirectXPixelFormat::B8G8R8A8UIntNormalized,
2,
size);
session_ = framePool_.CreateCaptureSession(captureItem_);
session_.IsCursorCaptureEnabled(false);
session_.IsBorderRequired(false);
return true;
}
bool WgcSession::initialize(HMONITOR monitor, int fps) {
fps_ = fps;
frameIntervalHns_ = 10000000LL / fps_;
if (!createD3DDevice()) return false;
winrtDevice_ = createWinRTDevice();
if (!winrtDevice_) {
std::cerr << "ERROR: Failed to create WinRT D3D device" << std::endl;
return false;
}
captureItem_ = createCaptureItemForMonitor(monitor);
return initializeWithItem(fps);
}
bool WgcSession::initialize(HWND hwnd, int fps) {
fps_ = fps;
frameIntervalHns_ = 10000000LL / fps_;
if (!createD3DDevice()) return false;
winrtDevice_ = createWinRTDevice();
if (!winrtDevice_) {
std::cerr << "ERROR: Failed to create WinRT D3D device" << std::endl;
return false;
}
captureItem_ = createCaptureItemForWindow(hwnd);
return initializeWithItem(fps);
}
void WgcSession::setFrameCallback(FrameCallback callback) {
frameCallback_ = std::move(callback);
}
bool WgcSession::startCapture() {
if (!session_ || !framePool_) return false;
capturing_ = true;
lastFrameTimeHns_ = 0;
frameArrivedRevoker_ = framePool_.FrameArrived(
winrt::auto_revoke,
[this](auto const& sender, auto const& args) {
onFrameArrived(sender, args);
});
session_.StartCapture();
return true;
}
void WgcSession::stopCapture() {
capturing_ = false;
frameArrivedRevoker_.revoke();
if (session_) {
session_.Close();
session_ = nullptr;
}
if (framePool_) {
framePool_.Close();
framePool_ = nullptr;
}
}
void WgcSession::onFrameArrived(
winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool const& sender,
winrt::Windows::Foundation::IInspectable const&) {
if (!capturing_) return;
auto frame = sender.TryGetNextFrame();
if (!frame) return;
auto timestamp = frame.SystemRelativeTime();
int64_t frameTimeHns = std::chrono::duration_cast<std::chrono::duration<int64_t, std::ratio<1, 10000000>>>(timestamp).count();
// Frame rate limiting: skip frames that arrive too soon
if (lastFrameTimeHns_ > 0 && (frameTimeHns - lastFrameTimeHns_) < (frameIntervalHns_ * 7 / 10)) {
frame.Close();
return;
}
lastFrameTimeHns_ = frameTimeHns;
auto surface = frame.Surface();
// Get the underlying D3D texture from the WinRT surface via COM interop
winrt::com_ptr<IDirect3DDxgiInterfaceAccess> access;
try {
access = surface.as<IDirect3DDxgiInterfaceAccess>();
} catch (...) {
frame.Close();
return;
}
ComPtr<ID3D11Texture2D> texture;
HRESULT hr = access->GetInterface(IID_PPV_ARGS(&texture));
if (SUCCEEDED(hr) && texture && frameCallback_) {
frameCallback_(texture.Get(), frameTimeHns);
}
frame.Close();
}
@@ -0,0 +1,62 @@
#pragma once
#include <windows.h>
#include <d3d11.h>
#include <dxgi1_2.h>
#include <wrl/client.h>
#include <winrt/Windows.Foundation.h>
#include <winrt/Windows.Graphics.Capture.h>
#include <winrt/Windows.Graphics.DirectX.h>
#include <winrt/Windows.Graphics.DirectX.Direct3D11.h>
#include <functional>
#include <atomic>
#include <string>
using Microsoft::WRL::ComPtr;
class WgcSession {
public:
using FrameCallback = std::function<void(ID3D11Texture2D*, int64_t timestampHns)>;
WgcSession();
~WgcSession();
bool initialize(HMONITOR monitor, int fps);
bool initialize(HWND hwnd, int fps);
void setFrameCallback(FrameCallback callback);
bool startCapture();
void stopCapture();
int captureWidth() const { return captureWidth_; }
int captureHeight() const { return captureHeight_; }
ID3D11Device* device() const { return d3dDevice_.Get(); }
ID3D11DeviceContext* context() const { return d3dContext_.Get(); }
private:
ComPtr<ID3D11Device> d3dDevice_;
ComPtr<ID3D11DeviceContext> d3dContext_;
winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice winrtDevice_{nullptr};
winrt::Windows::Graphics::Capture::GraphicsCaptureItem captureItem_{nullptr};
winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool framePool_{nullptr};
winrt::Windows::Graphics::Capture::GraphicsCaptureSession session_{nullptr};
winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool::FrameArrived_revoker frameArrivedRevoker_;
FrameCallback frameCallback_;
std::atomic<bool> capturing_{false};
int fps_ = 60;
int captureWidth_ = 0;
int captureHeight_ = 0;
int64_t frameIntervalHns_ = 0;
int64_t lastFrameTimeHns_ = 0;
bool createD3DDevice();
winrt::Windows::Graphics::DirectX::Direct3D11::IDirect3DDevice createWinRTDevice();
winrt::Windows::Graphics::Capture::GraphicsCaptureItem createCaptureItemForMonitor(HMONITOR monitor);
winrt::Windows::Graphics::Capture::GraphicsCaptureItem createCaptureItemForWindow(HWND hwnd);
bool initializeWithItem(int fps);
void onFrameArrived(
winrt::Windows::Graphics::Capture::Direct3D11CaptureFramePool const& sender,
winrt::Windows::Foundation::IInspectable const& args);
};
+2 -2
View File
@@ -3,7 +3,7 @@ import { mkdirSync, existsSync } from 'node:fs';
import path from 'node:path';
const projectRoot = process.cwd();
const sourceDir = path.join(projectRoot, 'electron', 'native', 'windows-capture');
const sourceDir = path.join(projectRoot, 'electron', 'native', 'wgc-capture');
const buildDir = path.join(sourceDir, 'build');
if (process.platform !== 'win32') {
@@ -81,7 +81,7 @@ try {
process.exit(1);
}
const exePath = path.join(buildDir, 'Release', 'windows-capture.exe');
const exePath = path.join(buildDir, 'Release', 'wgc-capture.exe');
if (existsSync(exePath)) {
console.log(`[build-windows-capture] Built successfully: ${exePath}`);
} else {