H264 Video Encoder for OpenGL

Though this long article has more lines than the encoder library itself, this is a very simple and easy to read and understand article. If you have read my other articles before, you will be comfortable to know I do not write complicated stuff.

Table of Contents


Video Encoder

  • Visual C++ 2015/2017
  • Windows 7/8/10

OpenGL Renderer

  • SDL2
  • SDL2 Image
  • GLEW
  • GLM
  • Tiny OBJ loader
  • Zlib

All the OpenGL Renderer required libraries are included in the repository. The focus is on Video Encoder.


I worked on this video encoder while writing my Windows Store App, Mandy Frenzy, a photo slideshow app for ladies. Right now, I am feeling burnt out so I am taking a short hiatus. Meanwhile, I am writing a series of short articles as a way to document this app. This video encoder is a header file only (H264Writer.h), based on Microsoft Media Foundation, not the old DirectShow as Microsoft did not expose the H264 and HEVC codec on DirectShow. It is tested on Windows 10. It should work fine on Windows 7/8 as well. Do let me know if you encounter any problem on those OSes. Windows 10 used to come bundled with HEVC codec. For unknown reasons, MS has taken it out in Windows 10 Fall Creators Update for the **new** Windows 10 installation and put it up for purchase for $1.50 in the Microsoft Store. In the section below, some screenshots will show encoding artifacts present in MS HEVC video.

What Are the Pros and Cons of This Encoder Over FFmpeg?

FFmpeg is GPL and so you may not concerned if you just want to encode the output of your personal renderer. For my freemium app, I prefer to steer clear of the licensing issues. How hobbyists usually encode their frames with FFmpeg is to save all the frames in HDD first which limits the number of frames and also directly impacted video duration that can be saved depending on the free HDD space. The extra step of saving and opening the files has negative impact of the encoding speed. Of course, tight integration with FFmpeg code may eliminate the frame saving part. On the other hand, this encoder reads RGB values from the framebuffer provided. The downside is it is not portable and only works on Windows 7/8/10.

3 Rendering Modes

The same OpenGL renderer can be compiled into 3 modes: normal OpenGL display mode, Video Encoder mode and Emscripten mode. The latter two’s code sections are respectively guarded by VIDEO_ENCODER and EMSCRIPTEN macros. You can, by all means, use your own renderer with the video encoder. The default OpenGL renderer is just provided to show a working demo.

The documentation is divided into 3 main sections. The first section is get the demo up and running and on how to modify the parameters. The second section is on how to integrate it with your OpenGL framework. The demo uses a renderer framework taught in Paul Varcholik’s OpenGL Essentials LiveLessons. The original source code used GLFW and is based on OpenGL 4.4: I converted his framework to use SDL and downgrade to OpenGL 2.0. The decision is based on the lowest denominator of what WebGL and Emscripten can support. A tutorial on how to integrate with DirectX will come later. In theory, this video encoder should integrate well with other graphics API like Vulkan, afterall, all it needs to be supplied with a video buffer and some synchronization in tandem to perform its work. The third section is on the explanation of the internals of the video encoder which you can skip if you are not into the encoder internals and implementation. And the last section explains the Emscripten part required to compile into asm.js or Webassembly.

Running the Demo

Spaceship Video

Spaceship Video on Youtube


All the required libraries are included in the repository. The required DLLs are copied automatically to the Release or Debug folder for Win32 post builds. x64 build is unbuildable due to the inability to find a x64 zlib lib/dll on the web; this is a linking problem that lies with the OpenGL renderer, not video encoder.

To see the OpenGL demo, open up SDL_App.sln in Visual Studio and build the SDL_App project.

To run the video encoding demo, open up H264SinkWriter.cpp and in the main function, modify the configFile, musicFile and videoFile to the paths on your machine. configFile is found in the $(SolutionDir)SDL_App folder. musicFile should be a mp3 file and if it is left blank, the final video shall not have music. videoFile is the encoded video. You can specify HVEC(aka H265) instead of H264 in the 4th parameter. HVEC is having some encoding issues where the colors bleed (See screenshots below.)

// This is the config file to be found in SDL_App project folder.
std::wstring configFile(L"D:\\GitHub\\video_encoder_for_ogl_dx\\SDL_App\\SDL_App\\config.txt");
// This is your music file
std::wstring musicFile(L"D:\\FMA.mp3");
// This is the video encoded output file.
std::wstring videoFile(L"C:\\Users\\shaov\\Documents\\video.mp4");

H264Writer writer(musicFile.c_str(), configFile.c_str(), videoFile.c_str(), VideoCodec::H264);
if (writer.IsValid())
    if (writer.Process())
        printf("Video written successfully!\n");
        return 0;
printf("Video write failed!\n");

The typical config.txt is to facilitate passing of information to OpenGL renderer, has nothing to do with video encoder. If your renderer can get the information about the video it is about to encode, then just pass a dummy config.txt. The contents of a typical config.txt is shown below:


Now the demo does not handle aspect ratio automatically and it always sticks with 4:3 ratio. If you enter anything which is 16:9, or wider than 4:3, in screen width and height, your video will look stretched. FPS entry is for the integer number of frames per second; there is no way to enter a decimal number like 29.7777.

The demo would only encode 5 seconds of the video. Change duration in RenderingScene::Draw function.

void RenderingScene::Draw(const SceneTime& gameTime)
    float grey = 35.0f / 255.0f;
    glClearColor(grey, grey, grey, 1.0f);



    static GLfloat start_time = gameTime.TotalGameTime();
    GLfloat elapsed_time = gameTime.TotalGameTime() - start_time;
    // During video encoding, only run for 5 seconds.
    if(elapsed_time > 5.0f)

You have to experiment to find out the optimal bitrate that can encode a good quality video.

enum class VideoCodec

// H264Writer constructor
H264Writer(const wchar_t* mp3_file, const wchar_t* src_file, const wchar_t* dest_file,
           VideoCodec codec, UINT32 bitrate = 4000000) :

To run the demo executable by itself, you need to copy the config.txt, Images and Models folders to the Release/Debug folder. The SDL2 DLLs would have already copied during post build.

By default, demo displays a rotating UFO saucer, to display other 3D model, just uncomment the other lines in CreateComponents().

void RenderingScene::CreateComponents()
    mCamera = std::unique_ptr(new
        FirstPersonCamera(*this, 45.0f, 1.0f / 0.75f, 0.01f, 100.0f));
    mServices.AddService(Camera::TypeIdClass(), mCamera.get());

        //mTexturedModelDemo = std::unique_ptr(new TexturedDemo(*this, *mCamera));

        //mDiffuseCube = std::unique_ptr(new DiffuseCube(*this, *mCamera,
        //    "Cube.obj.txt", "Cube.mtl.txt"));

        mUFOSpecularModel = std::unique_ptr(new SpecularModel(*this, *mCamera,
            "UFOSaucer3.jpg", "UFOSaucer.obj.txt.zip", "UFOSaucer.mtl.txt"));

        //mStarModel = std::unique_ptr(new StarModel(*this, *mCamera,
        //    glm::vec3(1.0f,1.0f,0.0f), "Star.obj.txt", "Star.mtl.txt"));
    catch (SceneException& e)
        char buf[1000];
        SPRINTF(buf, "ModelEffect exception:%s", e.what());

#ifdef __EMSCRIPTEN__

HEVC Artifacts

H264 Video of Mandy Frenzy

H264 Video on Youtube


HEVC Video (Artifacts) of Mandy Frenzy

HEVC Video on Youtube


As you can see, the sinewave artifacts in HEVC, not present in H264. Increasing the bitrate solves the problem at the expense of larger file size. By the way, the sinewave is rendered by triangles, not lines and not by fragment shaders. Reason being lines are usually implemented as 1 pixel wide in OpenGL ES 2.0. Using triangles allows me to control the width/height.

Integration with your OpenGL Framework

This section teaches the modification needed to integrate the video encoder into your renderer.

The fastest way to find all the encoding related code is to search for VIDEO_ENCODER macro in the source code. The encoder requires you to implement 2 global functions: check_config_file and encoder_start. See their declarations below. encoder_start is called in the worker thread.

extern int check_config_file(const wchar_t* file, int* width, int* height, int* fps);
extern int encoder_start(UINT** pixels, HANDLE evtRequest, HANDLE evtReply,
                         HANDLE evtExit, HANDLE evtVideoEnded, const WCHAR* szUrl);

They, in turn, called their DLL counterparts implemented in the Program.cpp.

int check_config_file(const wchar_t* file, int* width, int* height, int* fps)
    return ::check_project_file(file, width, height, fps);
int encoder_start(UINT** pixels, HANDLE evtRequest, HANDLE evtReply,
                  HANDLE evtExit, HANDLE evtVideoEnded, const WCHAR* szUrl)
    return ::encoder_main(pixels, evtRequest, evtReply, evtExit,
                          evtVideoEnded, szUrl);

check_project_file requires you to pass the screen width, height and FPS information from the file which is config.txt. In reality, you can just open any file which can provide you with this information, so implementation of check_project_file is unimportant. Let’s see how encoder_main is implemented.

SDL_APP_DLL_API int WINAPI check_project_file(const wchar_t* file, int* width,
                                              int* height, int* fps)
    ConfigSingleton config;
    const std::string afile = toAString(file);
    bool ret = config.LoadConfigFile(afile);
    if (ret)
        *width = config.GetScreenWidth();
        *height = config.GetScreenHeight();
        *fps = config.GetFPS();
    return ret;

SDL_APP_DLL_API int WINAPI encoder_main(UINT** pixels, HANDLE evtRequest,
                                        HANDLE evtReply, HANDLE evtExit,
                                        HANDLE evtVideoEnded, const WCHAR* szUrl)
        const std::string& config_file = toAString(szUrl);

        if (gConfigSingleton.IsLoadSuccess() == false)
            gLogger.Log("gConfigSingleton.IsLoadSuccess() failed! See log!");

            return 1;

        Texture::setScreenDim(gConfigSingleton.GetScreenWidth(), gConfigSingleton.GetScreenHeight());
        std::unique_ptr renderingScene(new RenderingScene(L"Photo Montage",
            gConfigSingleton.GetScreenWidth(), gConfigSingleton.GetScreenHeight()));
        renderingScene->initVideoEncoder(pixels, evtRequest, evtReply, evtExit,
            evtVideoEnded, gConfigSingleton.GetFPS());

    catch (SceneException& ex)
    catch (std::exception& ex)

    return 0;

encoder_main is very similar to WinMain except it calls initVideoEncoder function to hand over the parameters. This is how initVideoEncoder is implemented: it just zero initialized some members and saved the parameters inside its members. These HANDLE arguments are already initialized inside the H264Writer constructor. setTimeQuandant() is to set the time increment for every frame, for example, if FPS is 30, then the time increment should be 33.3, irregardless of actual time passed. You wouldn’t want varying time rate for your video encoding.

void Scene::initVideoEncoder(UINT** pixels, HANDLE evtRequest, HANDLE evtReply,
                             HANDLE evtExit, HANDLE evtVideoEnded, int fps)
    mTexture = 0;
    mRenderBuffer = 0;
    mDepthBuffer = 0;
    mMultisampleTexture = 0;
    mPixels = pixels;
    mPixelBuffer = nullptr;
    mEvtRequest = evtRequest;
    mEvtReply = evtReply;
    mEvtExit = evtExit;
    mEvtVideoEnded = evtVideoEnded;


This is how encoder_start is called in the worker thread started by H264Writer.

    H264Writer* pWriter = (H264Writer*)(pParam);
    return ::encoder_start((UINT**)(

In our OpenGL renderer, we need to create RenderBuffer for our renderer to draw on.

void Scene::CreateFBO()
    mPixelBuffer = new unsigned int[mScreenWidth*mScreenHeight];

    glGenTextures(1, &mMultisampleTexture);
    glBindTexture(GL_TEXTURE_2D_MULTISAMPLE, mMultisampleTexture);
    glTexImage2DMultisample(GL_TEXTURE_2D_MULTISAMPLE, 4,
                  GL_RGBA, mScreenWidth, mScreenHeight, GL_TRUE);

    glGenRenderbuffers(1, &mRenderBuffer);
    glBindRenderbuffer(GL_RENDERBUFFER, mRenderBuffer);
    glRenderbufferStorageMultisample(GL_RENDERBUFFER, 16, GL_RGBA8, mScreenWidth, mScreenHeight);
    glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, mRenderBuffer);

    // Create depth render buffer (This is optional)
    glGenRenderbuffers(1, &mDepthBuffer);
    glBindRenderbuffer(GL_RENDERBUFFER, mDepthBuffer);
    glRenderbufferStorageMultisample(GL_RENDERBUFFER, 16,
                       GL_DEPTH24_STENCIL8, mScreenWidth, mScreenHeight);
    glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, mDepthBuffer);
    glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, mDepthBuffer);

    GLuint mTexture = 0;
    glGenTextures(1, &mTexture);
    glBindTexture(GL_TEXTURE_2D, mTexture);
    unsigned int dim = determineMinDim(mScreenWidth, mScreenHeight);
    unsigned int* pixels = new unsigned int[dim * dim];
    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, dim, dim, 0, GL_RGBA, GL_UNSIGNED_BYTE, pixels);
    delete [] pixels;
    pixels = NULL;

    glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, mTexture, 0);

    // Enable multisampling

    if (GL_FRAMEBUFFER_COMPLETE == glCheckFramebufferStatus(GL_FRAMEBUFFER))
        OutputDebugStringA("FBO status: Complete!\n");
        OutputDebugStringA("FBO status: Not complete!\n");

determineMinDim() used in the above function is implemented in this way because we need a square texture to power of 2.

unsigned int Scene::determineMinDim(unsigned int width, unsigned int height)
    unsigned int dim = width;
    if (height > width)
        dim = height;

    unsigned int min_dim = 32;
    if (dim > 32 && dim  64 && dim  128 && dim  256 && dim  512 && dim  1024 && dim  2048 && dim <= 4096)
        min_dim = 4096;
        min_dim = 4096;

    return min_dim;

Lastly, we need to read the rendered buffer with glReadPixels. The code wait on mEvtRequest and mEvtExit. mEvtRequest is signaled when the encoding thread requests for new frame. Whereas the mEvtExit is signaled during exit. After copying the buffer to mPixels, mEvtReply is signaled to encoding thread that mPixels is ready.

void Scene::ReadBuffer(bool& quit)

    glReadPixels(0, 0, mScreenWidth, mScreenHeight, GL_BGRA_EXT, GL_UNSIGNED_BYTE, mPixelBuffer);

    HANDLE arrHandle[2];
    arrHandle[0] = mEvtRequest;
    arrHandle[1] = mEvtExit;

    DWORD dwEvt = WaitForMultipleObjects(2, arrHandle, FALSE, INFINITE);

    if (dwEvt == WAIT_OBJECT_0 + 1)
        quit = true;
    while (*mPixels == NULL) { Sleep(100); }

        memcpy(*mPixels, mPixelBuffer, mScreenWidth*mScreenHeight*sizeof(unsigned int));


This is how CreateFBO() and ReadBuffer() are called in my Render(). CreateFBO() and ReadBuffer() are empty when not compiled in VIDEO_ENCODER mode.

void Scene::Render(bool& quit)
    static bool all_init = false;
    if (all_init == false)
        if (IsAllReady())
            if (IsAllInitialized())
                all_init = true;
    if (all_init)
        if (mPaused == false)

How Is Video Encoder Written

Too long to be put in the blog post, read it at Github Readme

Running as asm.js on Web Browser

This section focuses mainly on asm.js aspect of OpenGL framework that comes bundled with the demo. If you are only interested in the video encoder, you can safely ignore this section. If you want to run your app on the web browser with the framework, this is the section for you. The design choice of framework is based on lowest technologies mainstream WebGL and web browser environment can work with. Since WebGL is a safe subset based on OpenGL 2.0 ES in such a way that WebGL calls can be translated to OpenGL 2.0 calls with little effort. Without coincidence, this is also the maximum OpenGL version the framework can support.

Change Your URL

Change your IP address/domain/port accordingly in SRC_FOLDER in the Program.cpp. SRC_FOLDER is used to download your assets from. Switch to http if you are not using https.

#ifdef __EMSCRIPTEN__
	#define SRC_FOLDER "https://localhost:44319/"
	#define SRC_FOLDER ".\\"

Recompile the Code as asm.js

Recompile the code for new URL to take effect, if you are on Windows 10 and are comfortable tinkering with Bash commands, enable Windows Subsystem for Linux and install Ubuntu from MS Store and install Emscripten. Run GNU make.

In the Makefile, change to -s WASM=0

make all

Recompile the Code as Webassembly

Remove or change to -s WASM=1 because default option is compiling to Webassembly when WASM is not specified. The reason I did not do so, because IIS do not recognise the wasm mime type added to it.

No Assimp Support

Assimp is not supported simply because there is no Emscripten port for Assimp. Instead, Tiny OBJ Loader is used to load simple 3D model in Wavefront OBJ format. OBJ file extensions ends in *.obj and *.mtl and I have modified the library to load in *.obj.txt and *.mtl.txt because I want to avoid adding new mime types in the web server.

Downloading Your Assets

For drawing anything on screen, derive your class from DrawableSceneComponent and **must** call DownloadFiles() in the constructor to download your assets from relative subfolders because asm.js must have all assets downloaded before the main rendering loop is run. DownloadFiles calls OpenFile on the files directly on a desktop app. If the code is compiled with EMSCRIPTEN macro defined, it will first download files and then open.

No Downloading of Shader

The shader code is inline with C++11 raw string literals to save downloading effort.

Reference Book

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this:
search previous next tag category expand menu location phone mail time cart zoom edit close