Show Me the Movie with DirectShow
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
The popularity of Web-based media is evidenced by an increasing number of Web pages that embed media players. Although these players are based mainly on Adobe's Flash and Microsoft's Silverlight technologies, some of them (with many more expected) are based on the JavaFX technology from Sun Microsystems.
JavaFX's media player infrastructure is provided by Java Media Components (JMC), which lets JavaFX applications play back Flash-based media. Furthermore, the Windows version of JMC leverages Microsoft's DirectShow technology to play back Windows Media Video (WMV) and other kinds of Windows-oriented media files.
This article shows you how to use DirectShow to create standalone media player applications for Windows. After presenting a primer that introduces a simple "Hello, world!"–style media player application to demonstrate some DirectShow fundamentals, the article reveals additional DirectShow capabilities while developing an increasingly sophisticated media player.
DirectShow Primer
DirectShow (formerly known as ActiveMovie) is an architecture for capturing and playing back multimedia streams on Windows platforms. It automatically detects and uses video and audio acceleration hardware when available, and is useful for creating audio-video capture applications, media players, video editors, file format converters (transcoders), and other media applications.
DirectShow uses filters and filter graphs to manage the complexity of different media sources, media formats, compressors/decompressors (codecs), and media hardware involved in media playback. Supported media formats include Advanced Systems Format (ASF), Motion Picture Experts Group (MPEG), Audio-Video Interleaved (AVI), MPEG Audio Layer-3 (MP3), WAV, and WMV.
A filter is a software component that serves as a DirectShow building block. It receives data as input, processes this data in some fashion, and outputs the processed data. For example, a filter might read AVI data from a local file on the hard drive; another filter might draw a sequence of decompressed AVI video frames onto the screen.
There are three kinds of filters:
- Source filters read data from files or other sources.
- Transform filters transform data in various ways.
- Rendering filters output data to the screen, speakers, or some other destination.
Applications can chain filters to other filters; the resulting chains are known as filter graphs. For example, a filter graph consisting of an AVI file source filter connected to an AVI decompressor filter, which is connected to a video renderer filter, can read an audio-less AVI file, decompress its video frames, and render these frames onto the screen.
Applications rely on the filter graph manager, a high-level component, to manage data flow through filter graphs. The application makes high-level API calls such as "Run" (move data through the graph) or "Stop" (stop the flow of data) on the filter graph manager, and can receive event notifications (such as the media has finished playing) from this component.
DirectShow's architecture is based on Microsoft's Component Object Model (COM), which means that its API consists of COM interfaces: IGraphBuilder is the filter graph manager object's main COM interface. Because programming to COM interfaces is somewhat more involved than regular Win32 API programming, this article's code is written in the C++ language, which simplifies this task.
Hello, DirectShow!
New programming languages are often introduced via simple applications that employ language syntax to output some kind of Hello, World! message. I've done something analogous for DirectShow by creating a simple hdshow (Hello, DirectShow!) media player application. Listing 1 presents this application's hdshow.cpp source code.
Listing 1 hdshow.cpp
// hdshow.cpp
// Hello, DirectShow!
#include <control.h>
#include <iostream.h>
#include <strmif.h>
#include <uuids.h>
void main (int argc, char *argv [])
{
if (argc != 2)
{
cerr << "usage: hdshow filespec\n";
return;
}
if (FAILED (CoInitialize (NULL)))
{
cerr << "hdshow: CoInitialize() failure\n";
return;
}
IGraphBuilder *pGraph;
HRESULT hr = CoCreateInstance (CLSID_FilterGraph, NULL,
CLSCTX_INPROC_SERVER, IID_IGraphBuilder,
(LPVOID *) &pGraph);
if (FAILED (hr))
{
cerr << "hdshow: CoCreateInstance() failure\n";
CoUninitialize ();
return;
}
IMediaControl *pControl;
hr = pGraph->QueryInterface (IID_IMediaControl, (LPVOID *) &pControl);
if (FAILED (hr))
{
cerr << "hdshow: unable to obtain IMediaControl interface\n";
pGraph->Release ();
CoUninitialize ();
return;
}
IMediaEvent *pEvent;
hr = pGraph->QueryInterface (IID_IMediaEvent, (LPVOID *) &pEvent);
if (FAILED (hr))
{
cerr << "hdshow: unable to obtain IMediaEvent interface\n";
pControl->Release ();
pGraph->Release ();
CoUninitialize ();
return;
}
WCHAR wPath [MAX_PATH];
MultiByteToWideChar (CP_ACP, 0, argv [1], -1, wPath, MAX_PATH);
hr = pGraph->RenderFile (wPath, NULL);
if (SUCCEEDED (hr))
{
hr = pControl->Run ();
if (SUCCEEDED (hr))
{
long evCode;
pEvent->WaitForCompletion (INFINITE, &evCode);
}
else
cerr << "hdshow: unable to run " << argv [1] << "\n";
}
else
cerr << "hdshow: unable to render " << argv [1] << "\n";
pEvent->Release ();
pControl->Release ();
pGraph->Release ();
CoUninitialize ();
}
Listing 1 first includes header files control.h (define the filter graph manager class's IMediaControl and IMediaEvent interfaces, and their IID_IMediaControl and IID_IMediaEvent interface IDs), iostream.h (define the cerr stream), strmif.h (define the manager's IGraphBuilder interface and its IID_IGraphBuilder interface ID), and uuids.h (define the manager's CLSID_FilterGraph class ID).
Listing 1 proceeds to define the main() entry-point function. This function's first task is to validate the number of passed command-line arguments, which must be two. By convention, the first argument is always the application's name (hdshow). However, the second argument is the path and name of the media file to play.
The main() function next invokes CoInitialize() to initialize the COM library. Assuming success (the FAILED() macro returns FALSE), main() invokes CoCreateInstance() to instantiate the filter graph manager, and obtain access to this object via the main IGraphBuilder interface.
If main() successfully obtains an IGraphBuilder, it uses this interface to locate the filter graph manager's IMediaControl and IMediaEvent interfaces. The former interface controls streaming via methods for starting and stopping the graph. The latter interface makes it possible to suspend hdshow until playback completes.
After invoking MultiByteToWideChar() to convert the command-line argument from ANSI format to a wide-character string, main() passes this string as an argument to IGraphBuilder's HRESULT RenderFile(LPCWSTR lpwstrFile, LPCWSTR lpwstrPlayList) method, which builds a filter graph that renders the file's media.
Assuming that RenderFile() succeeds (the file exists and the format is recognized), main() next invokes IMediaControl's HRESULT Run() method to run the filter graph. Behind the scenes, this method moves data through the filters and renders data as video and audio. When playback reaches the end of the stream, the graph continues to run although the filters are not streaming any more data.
Because the filter graph runs on a separate thread, main()'s thread immediately returns. It waits for playback to complete by invoking IMediaEvent's HRESULT WaitForCompletion(long msTimeout, long *pEvCode) method, passing INFINITE to msTimeout to block the calling thread until a completion event (whose EC_COMPLETE code is stored in *pEvCode) occurs.
Before terminating, main() releases its three COM interface pointers, and invokes CoUninitialize() to close the COM library.
I used Borland C++ 5.5.1 to compile hdshow.cpp and create an hdshow.exe executable. If you're unfamiliar with this free compiler, my Borland C++ 5.5.1 article provides instructions for downloading, installing, and using this tool.
Assuming that you install Borland C++ 5.5.1 and set up your environment as per the article's instructions, invoke the following command line to create hdshow.exe:
bcc32 -I..\..\include -L..\..\lib hdshow.cpp
Assuming that hdshow.exe is created, extract the hds.wmv file from this article's code archive and invoke the following command line to play this file's movie:
hdshow hds.wmv
DirectShow presents a window in which the movie plays out. Figure 1 shows one of the movie's frames.
)
Hello, DirectShow! rotates around its center.
The window closes automatically when the movie ends. You can close the window prior to the movie ending by clicking the title bar's X button.
Behind the scenes, RenderFile() inserts a File Source (Async) filter into the filter graph when hds.wmv is passed to this method. This filter is responsible for reading the contents of locally stored media files. However, if you pass the Uniform Resource Locator (URL) of an externally stored media file to RenderFile(), this method inserts a File Source (URL) filter into the filter graph.
For example, suppose you specify hdshow http://www.spacetelescope.org/static/archives/videos/medium_flash/astro_aq.flv at the command line. The RenderFile() method detects that you want to play URL-based media and inserts a File Source (URL) filter to stream the media over the network. Figure 2 shows one of this video's frames.
)
A frame from a video of the Sun, one of several copyright-free videos from the videos section of The European Homepage For The NASA/ESA Hubble Space Telescope website.