Video SDK Windows
AnthropicOfficialCreated by ZoomDistributed by Anthropic
- Skill
- Tool integration
- Software development
- MIT
- No issues found
AutoGineer summary
Written by AutoGineer from the skill’s own files, not by the author
This skill provides technical guidance for developers building custom video applications on Windows using the Zoom Video SDK. It covers session management, raw media capture and injection, screen sharing, and real-time communication features. The documentation includes architectural patterns, C++ integration steps, and critical implementation requirements such as the Windows message loop and delegate callback handling.
What it does
- Provides C++ integration patterns for Zoom video sessions.
- Explains raw audio and video data capture and injection.
- Details screen sharing and collaborative whiteboard implementation.
- Outlines requirements for cloud recording and live streaming.
- Offers troubleshooting guides for common build and runtime issues.
- Supports .NET application integration via C++/CLI wrappers.
When to use it
- When developing custom video conferencing or streaming applications on Windows.
- When needing to capture or inject raw audio and video frames.
- When implementing custom UI layouts for video sessions.
- When troubleshooting callback failures or session connection issues.
Advantages
- Provides direct access to raw YUV420 and PCM data streams.
- Includes official patterns for session lifecycle and event handling.
- Offers multiple rendering approaches via Canvas API or raw data pipes.
- Provides comprehensive sample applications for specific use cases.
Usage examples
- “How do I join a Zoom Video SDK session in C++?”
- The agent provides the initialization code, session context setup, and the critical Windows message loop required to process SDK callbacks.
- “Why are my SDK callbacks not firing?”
- The agent explains the necessity of a Windows message pump and provides the code snippet to integrate PeekMessage and DispatchMessage into the main application loop.
- “How do I render remote user video?”
- The agent compares the Canvas API and Raw Data Pipe approaches, providing code examples for subscribing to video streams using the Canvas API.
- “How do I implement the delegate interface?”
- The agent provides a template for the IZoomVideoSDKDelegate class and emphasizes the requirement to implement all 80+ virtual methods.
Install
Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/anthropics/video-sdk-windows/downloadUnzip it into the cross-client skills folder of your project:
mkdir -p .agents/skills && unzip video-sdk-windows.zip -d .agents/skills
Documentation by Zoom
Zoom Video SDK - Windows Development
Expert guidance for developing with the Zoom Video SDK on Windows. This SDK enables custom video applications, raw media capture/injection, cloud recording, live streaming, and real-time transcription on Windows platforms.
Official Documentation: https://developers.zoom.us/docs/video-sdk/windows/ API Reference: https://marketplacefront.zoom.us/sdk/custom/windows/ Sample Repository: https://github.com/zoom/videosdk-windows-rawdata-sample
Quick Links
New to Video SDK? Follow this path:
SDK Architecture Pattern - Universal 3-step pattern for ANY feature
Session Join Pattern - Complete working code to join a session
Windows Message Loop - CRITICAL: Fix callbacks not firing
Video Rendering - Display video with Canvas API
Reference:
Singleton Hierarchy - 5-level SDK navigation map
API Reference - Methods, error codes, timing rules
Delegate Methods - All 80+ callback methods
Sample Applications - Official samples guide
windows.md - Secondary overview doc (pointer-style)
SKILL.md - Complete documentation navigation
Having issues?
Callbacks not firing → Windows Message Loop
Build errors → Build Errors Guide
Video subscribe fails → Video Rendering (subscribe in
onUserVideoStatusChanged)Quick diagnostics → Common Issues
Building a Custom UI?
Canvas vs Raw Data - Choose your rendering approach
Raw Video Capture - YUV420 frame processing
SDK Overview
The Zoom Video SDK for Windows is a C++ library that provides:
Session Management: Join/leave video SDK sessions
Raw Data Access: Capture raw audio/video frames (YUV420, PCM)
Raw Data Injection: Send custom audio/video into sessions
Screen Sharing: Share screens or inject custom share sources
Cloud Recording: Record sessions to Zoom cloud
Live Streaming: Stream to RTMP endpoints (YouTube, etc.)
Chat & Commands: In-session messaging and command channels
Live Transcription: Real-time speech-to-text
Subsessions: Breakout room support
Whiteboard: Collaborative whiteboard features
Annotations: Screen share annotations
C# Integration: C++/CLI wrapper for .NET applications
Prerequisites
System Requirements
OS: Windows 10 (1903 or later) or Windows 11
Architecture: x64 (recommended), x86, or ARM64
Visual Studio: 2019 or 2022 (Community, Professional, or Enterprise)
Windows SDK: 10.0.19041.0 or later
.NET Framework: 4.8 or later (for C# applications)
Visual Studio Workloads
Install these workloads via Visual Studio Installer:
Desktop development with C++
MSVC v142 or v143 compiler
Windows 10/11 SDK
C++ CMake tools (optional)
.NET desktop development (for C# applications)
.NET Framework 4.8 targeting pack
C++/CLI support
Quick Start
C++ Application
#include <windows.h>
#include "zoom_video_sdk_api.h"
#include "zoom_video_sdk_interface.h"
#include "zoom_video_sdk_delegate_interface.h"
USING_ZOOM_VIDEO_SDK_NAMESPACE
// 1. Create SDK object
IZoomVideoSDK* video_sdk_obj = CreateZoomVideoSDKObj();
// 2. Initialize
ZoomVideoSDKInitParams init_params;
init_params.domain = L"https://zoom.us";
init_params.enableLog = true;
init_params.logFilePrefix = L"zoom_win_video";
init_params.videoRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;
init_params.shareRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;
init_params.audioRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;
ZoomVideoSDKErrors err = video_sdk_obj->initialize(init_params);
// 3. Add event listener
video_sdk_obj->addListener(myDelegate);
// 4. Join session (IMPORTANT: set audioOption.connect = false)
ZoomVideoSDKSessionContext session_context;
session_context.sessionName = L"my-session";
session_context.userName = L"Windows User";
session_context.token = L"your-jwt-token";
session_context.videoOption.localVideoOn = false;
session_context.audioOption.connect = false; // Connect audio after join
session_context.audioOption.mute = true;
IZoomVideoSDKSession* session = video_sdk_obj->joinSession(session_context);
// 5. CRITICAL: Add Windows message pump for callbacks to work
bool running = true;
while (running) {
// Process Windows messages (required for SDK callbacks)
MSG msg;
while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
TranslateMessage(&msg);
DispatchMessage(&msg);
}
// Your application logic here
Sleep(10);
}C# Application
using ZoomVideoSDK;
var sdkManager = new ZoomSDKManager();
sdkManager.Initialize();
sdkManager.JoinSession("my-session", "jwt-token", "User Name", "");Key Features
| Feature | Description |
|---|---|
| Session Management | Join, leave, and manage video sessions |
| Raw Video (YUV I420) | Capture and inject raw video frames |
| Raw Audio (PCM) | Capture and inject raw audio data |
| Screen Sharing | Share screens or custom content |
| Cloud Recording | Record sessions to Zoom cloud |
| Live Streaming | Stream to RTMP endpoints |
| Chat | Send/receive chat messages |
| Command Channel | Custom command messaging |
| Live Transcription | Real-time speech-to-text |
| C# Support | Full .NET Framework integration |
Sample Applications
Official Repository: https://github.com/zoom/videosdk-windows-rawdata-sample
| Sample | Description |
|---|---|
| VSDK_SkeletonDemo | Minimal session join - start here |
| VSDK_getRawVideo | Capture YUV420 video frames |
| VSDK_getRawAudio | Capture PCM audio |
| VSDK_sendRawVideo | Inject custom video (virtual camera) |
| VSDK_sendRawAudio | Inject custom audio (virtual mic) |
| VSDK_CloudRecording | Cloud recording control |
| VSDK_CommandChannel | Custom command messaging |
| VSDK_TranscriptionAndTranslation | Live captions |
See complete guide: Sample Applications Reference
Critical Gotchas and Best Practices
⚠️ CRITICAL: Windows Message Pump Required
The #1 issue that causes session joins to hang with no callbacks:
All Windows applications using the Zoom SDK MUST process Windows messages. The SDK uses Windows messages to deliver callbacks like onSessionJoin(), onError(), etc.
Problem: Without a message pump, joinSession() appears to succeed but callbacks never fire.
Solution: Add this to your main loop:
while (running) {
// REQUIRED: Process Windows messages
MSG msg;
while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
TranslateMessage(&msg);
DispatchMessage(&msg);
}
// Your application logic
Sleep(10);
}Applies to:
Console applications (no automatic message pump)
Custom main loops
Applications that don't use standard WinMain/WndProc
GUI applications using WinMain with standard message loop already have this.
Audio Connection Strategy
Best Practice: Set audioOption.connect = false when joining, then connect audio in the onSessionJoin() callback.
// During join
session_context.audioOption.connect = false; // Don't connect yet
session_context.audioOption.mute = true;
// In onSessionJoin() callback
void onSessionJoin() override {
IZoomVideoSDKAudioHelper* audioHelper = video_sdk_obj->getAudioHelper();
if (audioHelper) {
audioHelper->startAudio(); // Connect now
}
}Why: This pattern is used in all official Zoom samples. It separates session join from audio initialization for better reliability and error handling.
All Delegate Callbacks Must Be Implemented
The IZoomVideoSDKDelegate interface has 70+ pure virtual methods. ALL must be implemented, even if empty:
// Required even if you don't use them
void onProxyDetectComplete() override {}
void onUserWhiteboardShareStatusChanged(IZoomVideoSDKUser*, IZoomVideoSDKWhiteboardHelper*) override {}
// ... etcTip: Check the SDK version's zoom_video_sdk_delegate_interface.h for the complete list. The interface changes between SDK versions.
Memory Mode for Raw Data
Always use heap mode for raw data memory:
init_params.videoRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;
init_params.shareRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;
init_params.audioRawDataMemoryMode = ZoomVideoSDKRawDataMemoryModeHeap;Stack mode can cause issues with large video frames.
Thread Safety
SDK callbacks execute on SDK threads, not your main thread:
Don't perform heavy operations in callbacks
Don't call
cleanup()from within callbacksUse thread-safe queues for passing data to UI thread
Use mutexes when accessing shared state
Consult Official Samples First
When SDK behavior is unexpected, always check the official samples before troubleshooting:
Local samples:
C:\tempsdk\Zoom_VideoSDK_Windows_RawDataDemos\VSDK_SkeletonDemo\(simplest)C:\tempsdk\sdksamples\zoom-video-sdk-windows-2.4.12\Sample-Libs\x64\demo\
Official samples show correct patterns for:
Message pump implementation ✓
Audio connection strategy ✓
Error handling ✓
Memory management ✓
Video Rendering - Two Approaches
The Zoom SDK provides two different ways to render video. Choose based on your needs.
🎯 Canvas API (Recommended for Most Use Cases)
Best for: Standard applications, clean video quality, ease of implementation
The SDK renders video directly to your HWND. No YUV conversion needed.
// Subscribe to a user's video with Canvas API
IZoomVideoSDKCanvas* canvas = user->GetVideoCanvas();
if (canvas) {
ZoomVideoSDKErrors ret = canvas->subscribeWithView(
hwnd, // Your window handle
ZoomVideoSDKVideoAspect_PanAndScan, // Fit to window, may crop
ZoomVideoSDKResolution_Auto // Let SDK choose best resolution
);
if (ret == ZoomVideoSDKErrors_Success) {
// SDK is now rendering directly to your window!
}
}
// Unsubscribe when done
canvas->unSubscribeWithView(hwnd);Advantages:
✅ Best quality - SDK uses optimized, hardware-accelerated rendering
✅ No artifacts - Professional video quality
✅ Simple code - 3 lines to subscribe
✅ Better performance - No CPU-intensive YUV conversion
✅ Automatic scaling - SDK handles window resizing
✅ Aspect ratio - Built-in aspect ratio handling
Example from official .NET sample:
// Self video preview
IZoomVideoSDKCanvas* canvas = myself->GetVideoCanvas();
canvas->subscribeWithView(selfVideoHwnd, aspect, resolution);
// Remote user video
IZoomVideoSDKCanvas* remoteCanvas = remoteUser->GetVideoCanvas();
remoteCanvas->subscribeWithView(remoteVideoHwnd, aspect, resolution);Video Aspect Options:
ZoomVideoSDKVideoAspect_Original- Letterbox/pillarbox, no croppingZoomVideoSDKVideoAspect_FullFilled- Fill window, may crop edgesZoomVideoSDKVideoAspect_PanAndScan- Smart crop to fill windowZoomVideoSDKVideoAspect_LetterBox- Show full video with black bars
Resolution Options:
ZoomVideoSDKResolution_90PZoomVideoSDKResolution_180PZoomVideoSDKResolution_360P- Good balanceZoomVideoSDKResolution_720P- HD qualityZoomVideoSDKResolution_1080PZoomVideoSDKResolution_Auto- Let SDK decide (recommended)
🔧 Raw Data Pipe (Advanced Use Cases)
Best for: Custom video processing, effects, recording, computer vision
You receive raw YUV420 frames and handle rendering yourself.
// 1. Create a delegate to receive frames
class VideoRenderer : public IZoomVideoSDKRawDataPipeDelegate {
public:
void onRawDataFrameReceived(YUVRawDataI420* data) override {
int width = data->GetStreamWidth();
int height = data->GetStreamHeight();
char* yBuffer = data->GetYBuffer();
char* uBuffer = data->GetUBuffer();
char* vBuffer = data->GetVBuffer();
// Convert YUV420 to RGB and render
ConvertYUVToRGB(yBuffer, uBuffer, vBuffer, width, height);
RenderToWindow(rgbBuffer, width, height);
}
void onRawDataStatusChanged(RawDataStatus status) override {
// Handle video on/off
}
};
// 2. Subscribe to raw data
IZoomVideoSDKRawDataPipe* pipe = user->GetVideoPipe();
VideoRenderer* renderer = new VideoRenderer();
pipe->subscribe(ZoomVideoSDKResolution_720P, renderer);YUV420 to RGB Conversion (ITU-R BT.601):
void ConvertYUV420ToRGB(char* yBuffer, char* uBuffer, char* vBuffer,
int width, int height) {
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
int yIndex = y * width + x;
int uvIndex = (y / 2) * (width / 2) + (x / 2);
int Y = (unsigned char)yBuffer[yIndex];
int U = (unsigned char)uBuffer[uvIndex];
int V = (unsigned char)vBuffer[uvIndex];
// YUV to RGB conversion
int C = Y - 16;
int D = U - 128;
int E = V - 128;
int R = (298 * C + 409 * E + 128) >> 8;
int G = (298 * C - 100 * D - 208 * E + 128) >> 8;
int B = (298 * C + 516 * D + 128) >> 8;
// Clamp to [0, 255]
R = (R < 0) ? 0 : (R > 255) ? 255 : R;
G = (G < 0) ? 0 : (G > 255) ? 255 : G;
B = (B < 0) ? 0 : (B > 255) ? 255 : B;
// Store RGB (BGR format for Windows)
rgbBuffer[yIndex * 3 + 0] = (unsigned char)B;
rgbBuffer[yIndex * 3 + 1] = (unsigned char)G;
rgbBuffer[yIndex * 3 + 2] = (unsigned char)R;
}
}
}Render with GDI:
void RenderToWindow(unsigned char* rgbBuffer, int width, int height) {
HDC hdc = GetDC(hwnd);
BITMAPINFO bmi = {};
bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
bmi.bmiHeader.biWidth = width;
bmi.bmiHeader.biHeight = -height; // Negative for top-down
bmi.bmiHeader.biPlanes = 1;
bmi.bmiHeader.biBitCount = 24; // 24-bit RGB
bmi.bmiHeader.biCompression = BI_RGB;
RECT rect;
GetClientRect(hwnd, &rect);
StretchDIBits(hdc,
0, 0, rect.right, rect.bottom, // Destination
0, 0, width, height, // Source
rgbBuffer, &bmi,
DIB_RGB_COLORS, SRCCOPY);
ReleaseDC(hwnd, hdc);
}Disadvantages:
⚠️ CPU intensive - YUV conversion can cause frame drops
⚠️ Artifacts - Manual rendering may show tearing/artifacts
⚠️ Complex - More code to maintain
⚠️ Performance - Slower than Canvas API
Use Raw Data When:
Adding video filters/effects
Recording to custom formats
Computer vision processing
Custom compositing
Streaming to non-standard outputs
Self Video vs Remote Users
Self Video (your own camera):
Option A: Canvas API
IZoomVideoSDKSession* session = sdk->getSessionInfo();
IZoomVideoSDKUser* myself = session->getMyself();
IZoomVideoSDKCanvas* canvas = myself->GetVideoCanvas();
canvas->subscribeWithView(selfVideoHwnd, aspect, resolution);Option B: Video Preview (for self only)
IZoomVideoSDKVideoHelper* videoHelper = sdk->getVideoHelper();
videoHelper->startVideo(); // Start transmission
// For preview rendering
videoHelper->startVideoCanvasPreview(selfVideoHwnd, aspect, resolution);Remote Users (other participants):
Canvas API (recommended):
// In onUserJoin callback
void onUserJoin(IZoomVideoSDKUserHelper*, IVideoSDKVector<IZoomVideoSDKUser*>* userList) {
for (int i = 0; i < userList->GetCount(); i++) {
IZoomVideoSDKUser* user = userList->GetItem(i);
IZoomVideoSDKCanvas* canvas = user->GetVideoCanvas();
canvas->subscribeWithView(userVideoHwnd, aspect, resolution);
}
}Event-Driven Subscription Pattern
⚠️ CRITICAL: Video subscription must be event-driven and manual.
Key Events:
onSessionJoin- Subscribe to self videoonUserJoin- Subscribe to new remote usersonUserVideoStatusChanged- Re-subscribe when video turns on/offonUserLeave- Unsubscribe and cleanup
Complete Pattern:
class MainFrame : public IZoomVideoSDKDelegate {
private:
std::map<IZoomVideoSDKUser*, IZoomVideoSDKCanvas*> subscribedUsers_;
HWND videoWindow_;
public:
void onSessionJoin() override {
// Start your own video
IZoomVideoSDKVideoHelper* videoHelper = sdk->getVideoHelper();
videoHelper->startVideo();
// Subscribe to self video
IZoomVideoSDKUser* myself = sdk->getSessionInfo()->getMyself();
SubscribeToUser(myself);
}
void onUserJoin(IZoomVideoSDKUserHelper*,
IVideoSDKVector<IZoomVideoSDKUser*>* userList) override {
// Get current user to exclude self
IZoomVideoSDKUser* myself = sdk->getSessionInfo()->getMyself();
for (int i = 0; i < userList->GetCount(); i++) {
IZoomVideoSDKUser* user = userList->GetItem(i);
// IMPORTANT: Only subscribe to REMOTE users!
if (user != myself) {
SubscribeToUser(user);
}
}
}
void onUserVideoStatusChanged(IZoomVideoSDKVideoHelper*,
IVideoSDKVector<IZoomVideoSDKUser*>* userList) override {
IZoomVideoSDKUser* myself = sdk->getSessionInfo()->getMyself();
for (int i = 0; i < userList->GetCount(); i++) {
IZoomVideoSDKUser* user = userList->GetItem(i);
if (user != myself) {
// Re-subscribe when video status changes
SubscribeToUser(user);
}
}
}
void onUserLeave(IZoomVideoSDKUserHelper*,
IVideoSDKVector<IZoomVideoSDKUser*>* userList) override {
for (int i = 0; i < userList->GetCount(); i++) {
IZoomVideoSDKUser* user = userList->GetItem(i);
UnsubscribeFromUser(user);
}
}
void onSessionLeave() override {
// Cleanup all subscriptions
for (auto& pair : subscribedUsers_) {
IZoomVideoSDKCanvas* canvas = pair.second;
if (canvas) {
canvas->unSubscribeWithView(videoWindow_);
}
}
subscribedUsers_.clear();
}
private:
void SubscribeToUser(IZoomVideoSDKUser* user) {
if (!user || subscribedUsers_.find(user) != subscribedUsers_.end())
return;
IZoomVideoSDKCanvas* canvas = user->GetVideoCanvas();
if (canvas) {
ZoomVideoSDKErrors ret = canvas->subscribeWithView(
videoWindow_,
ZoomVideoSDKVideoAspect_PanAndScan,
ZoomVideoSDKResolution_Auto
);
if (ret == ZoomVideoSDKErrors_Success) {
subscribedUsers_[user] = canvas;
}
}
}
void UnsubscribeFromUser(IZoomVideoSDKUser* user) {
auto it = subscribedUsers_.find(user);
if (it != subscribedUsers_.end()) {
IZoomVideoSDKCanvas* canvas = it->second;
if (canvas) {
canvas->unSubscribeWithView(videoWindow_);
}
subscribedUsers_.erase(it);
}
}
};Key Points:
✅ Subscribe in response to events (onUserJoin, onUserVideoStatusChanged)
✅ Always exclude current user from remote subscriptions
✅ Unsubscribe on onUserLeave
✅ Clean up all subscriptions on onSessionLeave
✅ Track subscriptions in a map for lifecycle management
⚠️ Screen Share Subscription (DIFFERENT from Video!)
CRITICAL: Screen share subscription uses IZoomVideoSDKShareAction from the callback, NOT user->GetShareCanvas()!
// WRONG - This won't work for remote screen shares!
user->GetShareCanvas()->subscribeWithView(hwnd, ...);
// CORRECT - Use IZoomVideoSDKShareAction from onUserShareStatusChanged callback
void onUserShareStatusChanged(IZoomVideoSDKShareHelper* pShareHelper,
IZoomVideoSDKUser* pUser,
IZoomVideoSDKShareAction* pShareAction) {
if (!pShareAction) return;
ZoomVideoSDKShareStatus status = pShareAction->getShareStatus();
if (status == ZoomVideoSDKShareStatus_Start ||
status == ZoomVideoSDKShareStatus_Resume) {
// Subscribe to the share using Canvas API
IZoomVideoSDKCanvas* shareCanvas = pShareAction->getShareCanvas();
if (shareCanvas) {
shareCanvas->subscribeWithView(shareWindow_,
ZoomVideoSDKVideoAspect_Original);
}
}
else if (status == ZoomVideoSDKShareStatus_Stop) {
// Unsubscribe when share stops
IZoomVideoSDKCanvas* shareCanvas = pShareAction->getShareCanvas();
if (shareCanvas) {
shareCanvas->unSubscribeWithView(shareWindow_);
}
}
}Why is share different from video?
Video: Each user has one video stream → use
user->GetVideoCanvas()Share: A user can have multiple share actions (multi-share) → use
IZoomVideoSDKShareAction*from callbackThe
IZoomVideoSDKShareActionobject represents a specific share stream and contains the share status, type, and rendering interfaces
See also: Screen Share Subscription Example
Multi-User Video Layout
For multiple participants, you need one HWND per user:
// Create separate windows/panels for each user
HWND selfVideoWindow = CreateWindow(...); // Your video
HWND user1Window = CreateWindow(...); // User 1's video
HWND user2Window = CreateWindow(...); // User 2's video
// Subscribe each user to their own window
myself->GetVideoCanvas()->subscribeWithView(selfVideoWindow, ...);
user1->GetVideoCanvas()->subscribeWithView(user1Window, ...);
user2->GetVideoCanvas()->subscribeWithView(user2Window, ...);Layout Strategies:
Grid layout (2x2, 3x3)
Gallery view (scrollable)
Active speaker (large) + thumbnails
Picture-in-picture
Common Video Issues
| Issue | Cause | Solution |
|---|---|---|
| Video not showing | Not calling startVideo() | Call videoHelper->startVideo() in onSessionJoin |
| Artifacts/tearing | Using Raw Data Pipe | Switch to Canvas API |
| Poor performance | YUV conversion on UI thread | Use Canvas API or move conversion to worker thread |
| Video freezes | Not processing Windows messages | Add message pump to main loop |
| Can't see self | Subscribing to wrong user | Use session->getMyself() for self video |
| Seeing self in remote list | Not excluding self | Check if (user != myself) before subscribing |
Complete Documentation Library
This skill includes comprehensive guides organized by category:
Core Concepts (Start Here!)
SDK Architecture Pattern - Universal 3-step pattern for ANY feature
Singleton Hierarchy - 5-level navigation guide
Canvas vs Raw Data - Choose your rendering approach
Complete Examples
Session Join Pattern - JWT auth + session join with full code
Video Rendering - Canvas API video display
Screen Share Subscription - View remote screen shares (DIFFERENT from video!)
Raw Video Capture - YUV420 frame capture
Raw Audio Capture - PCM audio capture
Send Raw Video - Virtual camera (inject custom video)
Send Raw Audio - Virtual mic (inject custom audio)
Cloud Recording - Cloud recording control
Command Channel - Custom command messaging
Transcription - Live transcription/captions
UI Framework Integration
Win32 Native - Direct SDK usage with Canvas API (best performance)
WinForms (.NET) - C++/CLI wrapper + Raw Data Pipe
WPF (.NET) - C++/CLI wrapper + BitmapSource conversion
Production Quality Guidelines - Checklist and common issues
C++/CLI Wrapper Patterns (Wrapping ANY Native Library)
Complete Guide - 8 patterns for native→.NET interop
Pattern 1: Basic Structure - Project setup, class layout
Pattern 2: void* Pointers - Hide native types
Pattern 3: gcroot Callbacks - Native→Managed events
Pattern 4: IDisposable - Cleanup pattern
Pattern 5: Strings - String^ ↔ wstring/string
Pattern 6: Arrays - pin_ptr, Marshal::Copy
Pattern 7: Threading - UI thread dispatch
Pattern 8: LockBits - Fast image conversion
Common Errors - Troubleshooting
Troubleshooting
Windows Message Loop - CRITICAL: Why callbacks don't fire
Build Errors - SDK header dependency fixes
Common Issues - Quick diagnostics & error codes
References
API Reference - 5-level API hierarchy, methods, error codes
Delegate Methods - All 80+ callback methods
SKILL.md - Complete navigation guide
Most Critical Issues (From Real Debugging)
Callbacks not firing → Missing Windows message loop (99% of issues)
Video subscribe returns error 2 → Subscribing too early
See: Video Rendering - Subscribe in
onUserVideoStatusChanged
Abstract class errors → Missing virtual method implementations
See: Delegate Methods
Key Insight
Once you learn the 3-step pattern, you can implement ANY feature:
Get singleton → 2. Implement delegate → 3. Subscribe & use
Resources
Official Docs: https://developers.zoom.us/docs/video-sdk/windows/
API Reference: https://marketplacefront.zoom.us/sdk/custom/windows/
Dev Forum: https://devforum.zoom.us/
GitHub Samples: https://github.com/zoom/videosdk-windows-rawdata-sample
Working Sample:
C:\tempsdk\zoom-video-sdk-windows-sample\(complete implementation)
Need help? Start with SKILL.md for complete navigation.
Merged from video-sdk/windows/SKILL.md
Zoom Video SDK Windows - Complete Documentation Index
Quick Start Path
If you're new to the SDK, follow this order:
Overview → windows.md
Read the architecture pattern → concepts/sdk-architecture-pattern.md
Universal formula: Singleton → Delegate → Subscribe
Once you understand this, you can implement any feature
Fix build errors → troubleshooting/build-errors.md
SDK header dependencies
Required include order
Implement session join → examples/session-join-pattern.md
Complete working JWT + session join code
Fix callback issues → troubleshooting/windows-message-loop.md
CRITICAL: Why callbacks don't fire without Windows message loop
Implement video → examples/video-rendering.md
Canvas API (SDK-rendered) vs Raw Data Pipe
Troubleshoot any issues → troubleshooting/common-issues.md
Quick diagnostic checklist
Error code tables
Documentation Structure
video-sdk/windows/
├── SKILL.md # Main skill overview
├── SKILL.md # This file - navigation guide
├── windows.md # Secondary overview doc (pointer-style)
│
├── concepts/ # Core architectural patterns
│ ├── sdk-architecture-pattern.md # Universal formula for ANY feature
│ ├── singleton-hierarchy.md # 5-level navigation guide
│ └── canvas-vs-raw-data.md # SDK-rendered vs self-rendered choice
│
├── examples/ # Complete working code
│ ├── session-join-pattern.md # JWT auth + session join
│ ├── video-rendering.md # Canvas API video display
│ ├── screen-share-subscription.md # View remote screen shares
│ ├── raw-video-capture.md # YUV420 raw frame capture
│ ├── raw-audio-capture.md # PCM audio capture
│ ├── send-raw-video.md # Virtual camera (inject video)
│ ├── send-raw-audio.md # Virtual mic (inject audio)
│ ├── cloud-recording.md # Cloud recording control
│ ├── command-channel.md # Custom command messaging
│ ├── transcription.md # Live transcription/captions
│ └── dotnet-winforms/ # UI Framework integration
│ └── README.md # Win32, WinForms, WPF patterns
│ # C++/CLI wrapper patterns
│ # Production quality guidelines
│
├── troubleshooting/ # Problem solving guides
│ ├── windows-message-loop.md # CRITICAL - Why callbacks fail
│ ├── build-errors.md # Header dependency fixes
│ └── common-issues.md # Quick diagnostic workflow
│
└── references/ # Reference documentation
├── windows-reference.md # API hierarchy, methods, error codes
├── delegate-methods.md # All 80+ callback methods
└── samples.md # Official samples guideBy Use Case
I want to build a video app
SDK Architecture Pattern - Understand the pattern
Session Join Pattern - Join sessions
Video Rendering - Display video
Windows Message Loop - Fix callback issues
I'm getting build errors
Build Errors Guide - SDK header dependencies
Delegate Methods - Abstract class errors
Common Issues - Linker errors
I'm getting runtime errors
Windows Message Loop - Callbacks not firing
Common Issues - Error code tables
I want to view screen shares
Screen Share Subscription - DIFFERENT from video!
SDK Architecture Pattern - Event-driven pattern
Video Rendering - Compare with video subscription
I want to capture raw video/audio
Canvas vs Raw Data - Choose your approach
Raw Video Capture - YUV420 frame capture
Raw Audio Capture - PCM audio capture
SDK Architecture Pattern - Subscription pattern
I want to send custom video/audio (virtual camera/mic)
Send Raw Video - Inject custom video frames
Send Raw Audio - Inject custom audio
SDK Architecture Pattern - External source pattern
I want to record sessions
Cloud Recording - Start/stop cloud recording
API Reference - Recording helper methods
I want to use live transcription
Transcription - Enable live captions
Delegate Methods - Transcription callbacks
I want custom messaging between participants
Command Channel - Send custom commands
API Reference - Command channel methods
I want to build a Win32 native app
Win32 Integration - Direct SDK + Canvas API
Video Rendering - Canvas API patterns
Production Guidelines - Best practices
I want to build a WinForms (.NET) app
WinForms Integration - C++/CLI wrapper + Raw Data
C++/CLI Patterns - gcroot, Finalizer, LockBits
Production Guidelines - IDisposable, thread safety
I want to build a WPF (.NET) app
WPF Integration - C++/CLI + BitmapSource
Bitmap Conversion - Freeze(), Dispatcher
Production Guidelines - Performance optimization
I want to use C# / .NET Framework (general)
.NET Integration Overview - Complete C++/CLI wrapper guide
Raw Video Capture - YUV→RGB conversion patterns
Session Join Pattern - SDK initialization flow
I want to wrap ANY native C++ library for .NET
C++/CLI Wrapper Patterns - Complete 8-pattern guide
Pattern 1: Basic Structure - Project setup + class layout
Pattern 3: gcroot Callbacks - Native→Managed events
Pattern 4: IDisposable - Cleanup pattern
Common Errors - Troubleshooting
I want to implement a specific feature
SDK Architecture Pattern - START HERE!
Singleton Hierarchy - Navigate to the feature
API Reference - Method signatures
Most Critical Documents
1. SDK Architecture Pattern (MASTER DOCUMENT)
concepts/sdk-architecture-pattern.md
The universal 3-step pattern:
Get singleton (SDK, helpers, session, users)
Implement delegate (event callbacks)
Subscribe and use
2. Windows Message Loop (MOST COMMON ISSUE)
troubleshooting/windows-message-loop.md
99% of "callbacks not firing" issues are caused by missing Windows message loop.
3. Singleton Hierarchy (NAVIGATION MAP)
concepts/singleton-hierarchy.md
5-level deep navigation showing how to reach every feature.
Key Learnings
Critical Discoveries:
Windows Message Loop is MANDATORY
SDK uses Windows message pump for callbacks
Without it, callbacks are queued but never fire
Subscribe in onUserVideoStatusChanged, NOT onUserJoin
Video may not be ready when user joins
Wait for video status change callback
See: Video Rendering
Two Rendering Paths
Canvas API: SDK renders to your HWND (recommended)
Raw Data Pipe: You receive YUV frames (advanced)
See: Canvas vs Raw Data
Helpers Control YOUR Streams Only
videoHelper->startVideo()starts YOUR cameraTo see others, subscribe to their Canvas/Pipe
See: Singleton Hierarchy
UI Framework Integration Differs by Platform
Win32: Direct SDK, Canvas API (SDK renders to HWND) - best performance
WinForms: C++/CLI wrapper, Raw Data Pipe, YUV→Bitmap, InvokeRequired
WPF: Same wrapper + Bitmap→BitmapSource, Dispatcher, Freeze()
C++/CLI Wrapper Patterns (for ANY native library → .NET)
void*pointers - hide native types from managed headersgcroot<T^>- prevent GC from collecting managed references in native codeFinalizer + Destructor -
~Class()and!Class()for IDisposable cleanuppin_ptr+Marshal::Copy- array/buffer conversionLockBits- 100x faster than SetPixel for image manipulationThread marshaling - InvokeRequired (WinForms) / Dispatcher (WPF)
Audio Connection Timing
Set
audioOption.connect = falseduring joinCall
startAudio()inonSessionJoincallback
Quick Reference
"My code won't compile"
"Callbacks never fire"
"Video subscription returns error 2"
→ Video Rendering - Subscribe in onUserVideoStatusChanged
"Abstract class error"
"How do I implement [feature]?"
"How do I navigate to [controller]?"
"What error code means what?"
Document Version
Based on Zoom Video SDK for Windows v2.x
Happy coding!
Remember: The SDK Architecture Pattern is your key to unlocking the entire SDK. Read it first!
Operations
RUNBOOK.md - 5-minute preflight and debugging checklist.
Files
- ATTRIBUTION.mdAdded by AutoGineer990 B
- LICENSELicence file added by AutoGineer1.1 KiB
- RUNBOOK.md2.2 KiB
- SKILL.md40.2 KiB
- concepts/canvas-vs-raw-data.md10.1 KiB
- concepts/sdk-architecture-pattern.md8.6 KiB
- concepts/singleton-hierarchy.md22.2 KiB
- examples/cloud-recording.md8.9 KiB
- examples/command-channel.md8.9 KiB
- examples/dotnet-winforms/README.md38.0 KiB
- examples/raw-audio-capture.md7.2 KiB
- examples/raw-video-capture.md13.0 KiB
- examples/screen-share-subscription.md18.4 KiB
- examples/send-raw-audio.md9.1 KiB
- examples/send-raw-video.md10.7 KiB
- examples/session-join-pattern.md17.8 KiB
- examples/transcription.md11.7 KiB
- examples/video-rendering.md13.3 KiB
- references/delegate-methods.md32.1 KiB
- references/samples.md7.7 KiB
- references/windows-reference.md42.7 KiB
- troubleshooting/build-errors.md6.0 KiB
- troubleshooting/common-issues.md7.2 KiB
- troubleshooting/windows-message-loop.md6.9 KiB
- windows.md1.9 KiB
examples/command-channel.md
Command Channel
Complete working code for custom command messaging between participants.
Official Sample: VSDK_CommandChannel in videosdk-windows-rawdata-sample
Overview
The command channel enables custom data exchange between participants. Use cases:
Application-specific signaling
Game state synchronization
Custom control messages
Real-time collaboration data
┌─────────────────────────────────────────────────────────────────┐
│ COMMAND CHANNEL FLOW │
├─────────────────────────────────────────────────────────────────┤
│ Sender: │
│ getCmdChannel() → sendCommand() or sendCommandToAll() │
│ │
│ Receiver: │
│ onCommandReceived(sender, command) callback │
└─────────────────────────────────────────────────────────────────┘Limitations
| Limit | Value |
|---|---|
| Max message rate | 60 messages/second |
| Max message size | ~1KB recommended |
| Reliability | Best effort (not guaranteed) |
Note: Commands are not persisted - late joiners won't receive previous commands.
Complete Working Code
CommandHandler.h
#pragma once
#include <windows.h>
#include <string>
#include <functional>
#include "zoom_video_sdk_interface.h"
USING_ZOOM_VIDEO_SDK_NAMESPACE
class CommandHandler {
public:
CommandHandler(IZoomVideoSDK* sdk);
// Send commands
bool SendToAll(const std::wstring& command);
bool SendToUser(IZoomVideoSDKUser* user, const std::wstring& command);
// Connection status
bool IsConnected() const { return m_connected; }
// Callbacks from delegate
void OnCommandReceived(IZoomVideoSDKUser* sender, const zchar_t* command);
void OnConnectResult(bool success);
// Set message handler
using MessageCallback = std::function<void(IZoomVideoSDKUser*, const std::wstring&)>;
void SetMessageHandler(MessageCallback callback) { m_callback = callback; }
private:
IZoomVideoSDK* m_sdk;
IZoomVideoSDKCmdChannel* m_cmdChannel;
bool m_connected;
MessageCallback m_callback;
};CommandHandler.cpp
#include "CommandHandler.h"
#include <iostream>
CommandHandler::CommandHandler(IZoomVideoSDK* sdk)
: m_sdk(sdk)
, m_cmdChannel(nullptr)
, m_connected(false) {
}
bool CommandHandler::SendToAll(const std::wstring& command) {
if (!m_cmdChannel) {
m_cmdChannel = m_sdk->getCmdChannel();
}
if (!m_cmdChannel) {
std::cout << "Command channel not available" << std::endl;
return false;
}
ZoomVideoSDKErrors err = m_cmdChannel->sendCommand(nullptr, command.c_str());
if (err == ZoomVideoSDKErrors_Success) {
std::wcout << L"Sent to all: " << command << std::endl;
return true;
}
std::cout << "Send failed: " << err << std::endl;
return false;
}
bool CommandHandler::SendToUser(IZoomVideoSDKUser* user, const std::wstring& command) {
if (!user) return false;
if (!m_cmdChannel) {
m_cmdChannel = m_sdk->getCmdChannel();
}
if (!m_cmdChannel) {
return false;
}
ZoomVideoSDKErrors err = m_cmdChannel->sendCommand(user, command.c_str());
if (err == ZoomVideoSDKErrors_Success) {
std::wcout << L"Sent to " << user->getUserName()
<< L": " << command << std::endl;
return true;
}
std::cout << "Send failed: " << err << std::endl;
return false;
}
void CommandHandler::OnCommandReceived(IZoomVideoSDKUser* sender, const zchar_t* command) {
if (!sender || !command) return;
std::wstring cmdStr(command);
std::wcout << L"Command from " << sender->getUserName()
<< L": " << cmdStr << std::endl;
// Call user handler if set
if (m_callback) {
m_callback(sender, cmdStr);
}
}
void CommandHandler::OnConnectResult(bool success) {
m_connected = success;
std::cout << "Command channel " << (success ? "connected" : "failed") << std::endl;
}Using in Delegate
class MyDelegate : public IZoomVideoSDKDelegate {
private:
CommandHandler* m_cmdHandler;
public:
MyDelegate(IZoomVideoSDK* sdk) {
m_cmdHandler = new CommandHandler(sdk);
// Set message handler
m_cmdHandler->SetMessageHandler([this](IZoomVideoSDKUser* sender,
const std::wstring& cmd) {
HandleCommand(sender, cmd);
});
}
void onSessionJoin() override {
// Send hello to all participants
m_cmdHandler->SendToAll(L"hello");
}
void onCommandReceived(IZoomVideoSDKUser* sender, const zchar_t* strCmd) override {
m_cmdHandler->OnCommandReceived(sender, strCmd);
}
void onCommandChannelConnectResult(bool isSuccess) override {
m_cmdHandler->OnConnectResult(isSuccess);
}
private:
void HandleCommand(IZoomVideoSDKUser* sender, const std::wstring& cmd) {
// Parse and handle commands
if (cmd == L"ping") {
m_cmdHandler->SendToUser(sender, L"pong");
}
else if (cmd.find(L"action:") == 0) {
// Handle action command
std::wstring action = cmd.substr(7);
ProcessAction(action);
}
}
void ProcessAction(const std::wstring& action) {
std::wcout << L"Processing action: " << action << std::endl;
}
};JSON Command Pattern
For structured data, use JSON encoding:
#include <json/json.h>
// Send JSON command
void SendJsonCommand(CommandHandler* handler, const std::string& type,
const Json::Value& data) {
Json::Value root;
root["type"] = type;
root["data"] = data;
Json::StreamWriterBuilder builder;
std::string jsonStr = Json::writeString(builder, root);
std::wstring wideStr(jsonStr.begin(), jsonStr.end());
handler->SendToAll(wideStr);
}
// Receive and parse JSON
void HandleJsonCommand(const std::wstring& cmd) {
std::string narrowStr(cmd.begin(), cmd.end());
Json::Value root;
Json::CharReaderBuilder builder;
std::istringstream stream(narrowStr);
if (Json::parseFromStream(builder, stream, &root, nullptr)) {
std::string type = root["type"].asString();
Json::Value data = root["data"];
if (type == "position") {
int x = data["x"].asInt();
int y = data["y"].asInt();
// Handle position update
}
}
}
// Usage
Json::Value posData;
posData["x"] = 100;
posData["y"] = 200;
SendJsonCommand(cmdHandler, "position", posData);IZoomVideoSDKCmdChannel Methods
| Method | Returns | Description |
|---|---|---|
sendCommand(user, cmd) | ZoomVideoSDKErrors | Send to specific user (NULL = all) |
Rate Limiting
The command channel is limited to 60 messages/second:
class RateLimitedSender {
std::chrono::steady_clock::time_point m_lastSend;
static const int MIN_INTERVAL_MS = 17; // ~60/sec
public:
bool SendWithRateLimit(CommandHandler* handler, const std::wstring& cmd) {
auto now = std::chrono::steady_clock::now();
auto elapsed = std::chrono::duration_cast<std::chrono::milliseconds>(
now - m_lastSend).count();
if (elapsed < MIN_INTERVAL_MS) {
std::this_thread::sleep_for(
std::chrono::milliseconds(MIN_INTERVAL_MS - elapsed));
}
m_lastSend = std::chrono::steady_clock::now();
return handler->SendToAll(cmd);
}
};Common Issues
Commands Not Received
Cause: Channel not connected
Fix: Wait for onCommandChannelConnectResult(true):
void onCommandChannelConnectResult(bool isSuccess) override {
if (isSuccess) {
// Now safe to send commands
}
}Error 8 (Too Frequent)
Cause: Exceeding 60 messages/second limit
Fix: Add rate limiting (see above)
Unicode Issues
Cause: Encoding mismatch
Fix: Use std::wstring consistently:
m_cmdChannel->sendCommand(user, L"message"); // Wide string literalRelated Documentation
Session Join Pattern - Session setup
Delegate Methods - Command callbacks
API Reference - Method signatures
Requirements and compatibility
Written for: Claude Code, Claude apps
Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta
“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.
Needs: Claude Code frontmatter
Security
No issues found by our static scan· No issues found
A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.
Attribution and licence
- Created by
- Zoom
- Publisher
- Anthropic (Official)
- Distributed by
- Anthropic
- Source
- anthropics/knowledge-work-plugins · partner-built/zoom-plugin/skills/video-sdk/windows · commit e050c2e
- Licence
- MIT License (redistribution allowed)
- LICENSE (added by AutoGineer from partner-built/zoom-plugin/LICENSE)
- ATTRIBUTION.md (written by AutoGineer)
- Synced
AutoGineer is not affiliated with or endorsed by Zoom or Anthropic.
Download ZIP
The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download video-sdk-windows.zip