DirectX.Capture Class Library

Brian Low,              29 Mar 2008                                      Public Domain

1

2

3

4

5

4.94/5 - 222 votes

5 removed

μ 4.86, σ_a 0.93 [?]

Rate this: vote 1vote 2vote 3vote 4vote 5

Please Sign up or sign in to vote.

A .NET class library for capturing video and audio to AVI files.

• Download source files - 140.95 KB
• Download source files with demo project - 193.98 KB

Introduction

This article presents a class library for capturing audio and video to AVI files in .NET. Some of the features of this library:

• List and select hardware devices
• Access to common audio and video settings (e.g. frame rate, size)
• Support audio and video compression codecs
• Support video preview
• Support TV tuners
• Support crossbars and audio mixers
• Retrieve device capabilities
• Show property pages exposed by drivers
• MSDN-style documentation included

Using the Code

The Capture class is the core of this library. Here is a simple example:

Hide    Copy Code

// Remember to add a reference to DirectX.Capture.dll
using DirectX.Capture

// Capture using the first video
// and audio devices available
Capture capture = new Capture( Filters.VideoInputDevices[0],
                               Filters.AudioInputDevices[0] );

// Start capturing
capture.Start();

// Stop capturing
capture.Stop();

Remember to add a reference in your project to DirectX.Capture.dll. This DLL requires DShowNET.dll, so make sure they are both in the same directory. Once you add the reference, Visual Studio .NET should take care of the copying for you.

This example will capture video and audio using the first video and audio devices installed on the system. To capture video only, pass a null as the second parameter to the constructor.

The class is initialized to a valid temporary file in the Windows temp folder. To capture to a different file, set the Capture.Filename property before you begin capturing.

A Second Example

This next example shows how to change video and audio settings. Properties such as Capture.FrameRate and Capture.AudioSampleSize allow you to programmatically adjust the capture. Use Capture.VideoCaps and Capture.AudioCaps to determine valid values for these properties.

Hide    Copy Code

Capture capture = new Capture( Filters.VideoInputDevices[0],
                               Filters.AudioInputDevices[1] );

capture.VideoCompressor = Filters.VideoCompressors[0];
capture.AudioCompressor = Filters.AudioCompressors[0];

capture.FrameRate = 29.997;                 // NTSC
capture.FrameSize = new Size( 640, 480 );   // 640x480
capture.AudioSamplingRate = 44100;          // 44.1 kHz
capture.AudioSampleSize = 16;               // 16-bit
capture.AudioChannels = 1;                  // Mono

capture.Filename = "C:\MyVideo.avi";

capture.Start();
...
capture.Stop();

The example above also shows the use of video and audio compressors. In most cases, you will want to use compressors. Uncompressed video can easily consume over 1GB of disk space per minute. Whenever possible, set the Capture.VideoCompressor and Capture.AudioCompressor properties as early as possible. Changing them requires the internal filter graph to be rebuilt which often causes most of the other properties to be reset to default values.

Behind the Scenes

This project uses 100% DirectShow to capture video. Once a capture is started, DirectShow spawns another thread and handles retrieving/moving all the video and audio data itself. That means you should be able to capture at the same speed and quality as an application written in C.

DirectShow is implemented as a set of COM components and we use .NET Interop to access them. The pioneering work on this was done by NETMaster with the DShowNET project. This Capture library uses DShowNET for the interop layer with only a few extensions. This is the DShowNET.dll mentioned earlier.

Sitting on top of all of this is the Capture class library. The center of any DirectShow app is the filter graph and the filter graph manager. For a good overview, see The Filter Graph and Its Components from the MSDN.

The Least Work Possible

The library tries at all times to do the least amount of work possible. The problem is: DirectShow is very flexible, but has few firm standards for driver developers and I have limited hardware to test with. As a result, the class tries to avoid doing any work that may not be necessary, hopefully avoiding potential incompatibilities in the process.

One example is video preview. You can start and stop preview with:

Hide    Copy Code

// Start preview
capture.PreviewWindow = myPanelControl;

// Stop preview
capture.PreviewWindow = null;

Hopefully this is simple to use. Internally, DirectShow does a lot of work: add required upstream filters for WDM devices, search for preview pins, use the Overlay Manager for video ports (hardware overlays), insert SmartTee filters when a separate preview pin is not available and more. Instead of rendering the preview stream as soon as the class is created, the class waits until the PreviewWindow property is set.

For developers who don‘t need preview, none of this work will ever be done. That means your application is more likely to work on a wider range of hardware. For developers that do need preview, this makes it easier to locate the cause of the problem and fix it or handle it gracefully.

Performance Tips

Many of the properties on the Capture class are retrieved directly from the underlying DirectShow COM components. If you need to refer to the property repeatedly in a block of code, take a copy of the value and use your copy.

Hide    Copy Code

// AudioSampleSize is retrieved from DirectShow each iteration
for ( int c = 0; c < 32; c++ )
{
    if ( c == capture.AudioSampleSize )
        MessageBox.Show( "Found!" );
}

// A faster solution
int x = capture.AudioSampleSize;
for ( int c = 0; c < 32; c++ )
{
    if ( c == x )
        MessageBox.Show( "Found!" );
}

Why doesn‘t the class simply cache the value internally? We don‘t know when the filter (device driver) will change this value, so we have to retrieve the value every time. This means you will always get the real value of the property.

Credits

The DirectShow interop layer was developed by NETMaster in the DShowNET project. The MDSN-style documentation was generated from the source code using nDoc.

Troubleshooting

I have tested this with an Asus v7700 (NVidia GeForce2, reference drivers) and my onboard sound card. I can‘t guarantee any other hardware will work. However, I expect most video capture cards and sound cards will work. You may have trouble with TV Tuner cards and DV devices (Firewire camcorders) though they should be solvable.

Try the AMCap sample from the DirectX SDK (DX9\Samples\C++\DirectShow\Bin\AMCap.exe) or Virtual VCR, a free DirectShow capture application.

This class library uses COM Interop to access the full capabilities of DirectShow, so if there is another application that can successfully use a hardware device then it should be possible to modify this class library to use the device. Please post your experiences, good or bad, in the forum below.

User Enhancements

The following enhancements have been posted to the discussion board:

• Frequency Overrides for the Tuner class by fdaupias
• Radio Tuning for the Tuner class by dauboro

Thanks to fdaupias and dauboro for their submissions. I have not had time to post a tested, updated version with these enhancements. If anyone wants to make an updated download zip, mail it to me and I will added it to this page. Keep the enhancements coming.

DirectX.Capture Wiki

A Wiki for this project is available here. This Wiki can be edited by anyone, no registration is required. I hope this Wiki will allow interested users to more easily collaborate on this project. New versions, enhancements, tips and tricks can be posted on the Wiki.

License

This article, along with any associated source code and files, is licensed under A Public Domain dedication