The Don's Craptastic World » virtualdub

Intel Core i7-2600K and encoding with VirtualDub

Don Salva — Wed, 14 Dec 2011 16:44:53 +0000

A couple weeks ago I retired my old 3GHz overclocked Q6600 and upgraded to Intel's new flagship CPU i7 2600k. Now before I start writing up my very short opinion and review, those are the settings I still use in VirtualDub. Virtually nothing has changed. Even after more than 2 years, those settings are still the best in terms of "filesize:video quality"-ratio.

Encoding speed is measured in (average) frames per second. The more, the merrier. It's also important to know that there are, what I like to call, "motion-heavy" and "motion-light" scenes.

This is, like the name suggests, a scene in-game (or anywhere for that matter, but I'm talking video gameplay specifically here) in where there's either very much motion / moving happening on the screen or the opposite. Why is that important you ask?

Motion-heavy scenes take much longer to encode than motion-light ones, at least that was my experience in those 2+ years since. I might of course be totally wrong about this, and unfortunately I don't have hard statistics to show for, i.e. benchmarks while encoding. Just my observation, unfortunately.

This is of course explained as if you were 5 years old. There are much, much more complex processes going on during encoding which I won't get into because this article isn't about that. Please also bear in mind that I'm by no means an expert in video encoding. I do it as a private hobby and I say what I have observed.

Right, moving on then. A standard 10 min. uncompressed gameplay video recorded in 1920*1080 with FRAPS is about 20GB large.

When I was using my old, overclocked 3GHz Intel Quad Core Q6600, with the settings linked above, I encoded with about 8fps in motion heavy videos and about 13fps in motion light videos.

With my i7-2600k, clocked at stock speed, I average at 16fps in motion heavy videos and at 25fps in motion light videos. That's an increase of 50% in motion heavy scenes and 52% in motion light scenes.

For example my recent video was about 24GB uncompressed. And compared to action games there's not very much moving/motion happening in those 11 minutes. This video took me, with above linked settings, 15 min. to fully encode at average 24fps. The final size was about 200mb.

VirtualDub makes use of HyperThreading, while encoding, this is how my CPU load looks like. The blue bars are HT, the green ones cores.

Was the purchase worth it (for me)? Absolutely. If you, like me, are encoding quite a lot, the 2600k is a good choice. It brought me a huge boost in encoding speed and will save me a lot of time down the road. Not to mention it will encode even faster when I overclock my 2600k to 4GHz and beyond.

Unfortunately I've no access to a 2500k, so I can't compare the 2600k to the 2500k and see for myself how much better or worse the 2500k performs under the same set of circumstances. If I'd hazard a guess, I'd say not by a whole lot.

The 2500k is essentially a 2600k without the HyperThreading, and I don't quite believe HyperThreading alone makes up for the huge performance jump from Q6600 to 2600k. So if you don't care about a few dollars, get the 2600k. If budget is a priority, get the 2500k. Or wait for Ivy-Bridge CPUs, which will come out next year.

VirtualDub: How to use mkv, mp4, flv, mov and other file formats with VirtualDub

Don Salva — Sun, 03 Jul 2011 13:35:19 +0000

A lot of people keep asking me how to use VirtualDub with other files formats besides .avi. Like mkv, flv, mp4 or even .mov.

VirtualDub, as stand-alone, cannot handle mkv, mp4, flv, mov or any other files besides avi. You need a plugin or plugins to load anything else besides avi. Mind you, VirtualDub still only can and will save in avi!

There are a couple plugins that can handle loading different files formats. For me personally only 2 come to mind:

There's my personal favorite: Ffmpeg Input Driver which can handle 31 different file formats according to the plugin developer. Among them are the popular mkv, mp4, flv, mov, rm, rvmb, wmv and ts. The plugin is based, as the name may suggest, of the ffmpeg libraries. And if something doesn't work, remember, this is a plugin still in early development stage as of 03.07.2011.

Then there's fccHandler, which can handle mpeg-2, flv, wmv, fli, flc, mkv. However fccHandler also offers the AC-3 ACM codec, which I use for sound in my videos.

Download whichever plugin you prefer, and unpack it into your VirtualDub plugin folder.

Remember however that those plugins only allow you to load said file formats. You cannot save those loaded files in mp4, mov, flv, only in avi.

VirtualDub: Use ffdshow filters (resize, sharpen, wartermark) in VirtualDub

Don Salva — Wed, 09 Feb 2011 10:05:25 +0000

In the guides here and here I talk about using the VirtualDub internal filters, namely the resize and sharpen filters. But you can also use external filters. I use the ffdshow filters.

What is ffdshow, you ask? Here's the quote from the ffdshow sourceforge page

ffdshow tryouts is a DirectShow and Video for Windows codec with support for a wide range of audio and video formats, such as Xvid, DivX, and H.264. It includes a powerful filter set that can enhance the video quality - with filters for resizing, deinterlacing, and displaying subtitles - as well as audio quality through normalization, down-/upmixing, and resampling.

First of all, you'll need to download the latest ffdshow codec. Although it's "safer" to always use the latest stable builds, in this case however I'd recommend always using the latest SVN builds. Because they are updated way more often including the filters. The latest SVN build, as of the time writing this article was 2011-01-19.

Even the SVN builds are subdivided in 3 areas:

Start installing the ffdshow codec now. You'll be asked what you want to install. This is your choice:

After that leave the rest of the settings as they are and continue finishing the codec installation. After the installation a file named ffvdub.vdf should be inside your VirtualDub/Plugins folder:

Now we can start using the ffdshow filter in VirtualDub. Start up VirtualDub, open a video file, then add the filter:

Now the important part: VirtualDub processes the selected filters in their order. Meaning concurrent. This is very important because it affects your quality especially in connection with watermarks.

Say such as if we were using the filters in this order: 1) waterwark, 2) sharpen, 3) resize. What happens is this: First a watermark is slapped on a frame, then this frame including the watermark is sharpened and after that the sharpened frame including the watermark is resized. Now this is not an idea. Why?

Imagine you have a watermark 100*100 pixel. This image is first slapped on each frame, then it is sharpened and then it is resized. You don't want that. You want the watermark to maintain its original quality and size. So the ideal filter order would be:

Resize the frame, sharpen the frame and insert original, unedited watermark. Let's take a look at the selected filters.

ffdshow filter: Resize

It's pretty much straightforward. Enable resize, choose "Specify horizontal and vertical size" and enter the numbers you want to resize it to. It's 1280*720 for 720p HD and 1920*1080 for 1080p HD.

If you're resizing down or up from a 4:3 resolution, remember to enable "Keep original aspect ratio" under Aspect ratio. Otherwise the end-result will look over-stretched/-squeezed.

If you're resizing down or up from a 16:10 resolution, you could ignore this option. The end-result won't look too bad. But check it out yourself to make sure!

ffdshow filter: Sharpen

Sharpen is a bit trickier, and a lot different from VirtualDub's internal one. In fact there are 5 different sharpening algorithms, I haven't experience with them all, but I've made best experiences with "msharpen". Enable "high quality".

Sadly I can't give you ballpark figures like in the previous guides because this filter works very different. I'll do my best to try to explain how it works.

There are 2 sliders: Strength (1-127) and Threshold (0-255). Strength dictates how strong it should sharpen. Threshold dictates how much of the frame should be sharpened. A Threshold of 0 means everything (every pixel) in the frame will be sharpened, a Threshold of 255 sharpens nothing in the frame.

In order for you to see how much of a frame is sharpened enable "only mask" and press "Preview", play with the slider "Threshold" and set it to your liking. All that "white" in your Preview will be sharpened. Once you've set your Threshold to your liking don't forget to turn off "only mask"! Now play with the slider Strength and set to your liking. Each video needs a different setting I've concluded so far.

This is how the different msharp settings look like: Sharpen filter off, Sharpen Strength 127/Threshold 255, Sharpen Strength 127/Threshold 0, Sharpen Strength 35/Threshold 20, Sharpen Strength 30/Threshold 3.

Sharpen filter with "only mask" activated: Threshold 255, Threshold 0, Threshold 30, Threshold 15, Threshold 7, Threshold 3, Threshold 1.

ffdshow filter: Bitmap overlay (Watermark)

First thing fist: The watermark image needs to be a bitmap (.bmp). Everything else is pretty straightforward. Position your watermark with the sliders "Horizontal/Vertical position". Choose the mode of watermark inclusion and its Opacity. Preview your settings.

VirtualDub: How to prevent asynchronous sound

Don Salva — Tue, 24 Aug 2010 18:15:45 +0000

According to some comments from here, here and in some of the Emails I get people sometime experience asynchronous sound. As I have speculated here, this is most likely due the CPU being overwhelmed with the whole encoding process. Let me try to explain, and bear in mind, this is just speculation based on nothing but my own suspicions.

When you set the threads in the x264 codec to fully utilize your CPU, meaning all cores, and then some, you make sure the CPU is working at its maximum capacity just for video processing.

If you then add audio to the pile, the CPU just hasn't got any resources to spare, and if it does, it does a miserable job, thus resulting in lag which translates in asynchronous sound.

If that is the case, lowering the numbers of threads in the x264 codec should fix it. This way you can have both audio and video encoding in one go. I haven't tried this method, because I employ a different solution.

How I do it:

I process video and audio separately. First I encode/transcode the video, leaving the audio in its original form. Then I encode the audio and overwrite it.

Let say we have recorded a video and named this video myvideo_unedited.avi .

Transcode/Encode your myvideo_unedited.avi according to this or this BUT leave "Audio" at "Direct Stream copy". DO NOT change it like suggested in the aforementioned links and name this new video for instance transcoded-myvideo_RAWAUDIO.avi.

Now open transcoded-myvideo_RAWAUDIO.avi and set your "Video" to "Direct Stream copy". In "Audio" you set it to "Full Processing Mode". Go on and set your "Compression", "Volume", and other settings to your liking and save this new sound via "File"->"Save WAV...". The WAV-file is called the same as the video, in this case it'll be transcoded-myvideo_RAWAUDIO.wav.

Now set your "Audio" to "Direct Stream copy", and click "Audio from another file..." and select the new transcoded-myvideo_RAWAUDIO.wav and save this new video with its new audio via "File"->"Save as AVI..." and name it for example myvideo_final.avi.

You're surely asking why I'm doing this stupid extra work, when I can just lower the threads and be possibly done with it. The answer is transcoding/encoding speed. Lowering the threads in the x264 codec translates into longer transcoding time. Processing video needs much, much more time, than processing sound does.

Say for example, if I were to encode a 15 min. video, and I'd lower the threads just so I can process sound and video in one go, the whole process might take me maybe 1 hour, 30 min. If I encode video first, and leave the sound alone, the process takes me about 1 hour to complete. Then I process the sound alone which takes about 10 seconds. After that I merge both and I'm done in 1 hour and 5 min. That's 25 minutes shaved off.

Another important notice: Sometimes the sound is asynchronous when you play it in your vlc or any other player. However after you upload your video to YouTube the sound magically seems to be synchronous again. I haven't got a clue why that is but it happens sometimes.

VirtualDub: How to split and merge video files

Don Salva — Tue, 24 Aug 2010 17:27:20 +0000

Folks have been asking me here, here and on my YouTube channel how to split a single movie file into multiple ones, and how to merge multiple ones into a single movie.

I made 2 How-to clips way back then, but somehow forgot to include them in my previous tutorials, so here they are:

How To: Split a single video file into multiple video files with VirtualDub

Click here to view the embedded video.

How To: Add multiple video files into one single one with VirtualDub

Click here to view the embedded video.

VirtualDub: How to watermark your videos

Don Salva — Sat, 21 Aug 2010 09:07:48 +0000

There are many folks out there who like to watermark their videos in all possible kind of ways. Some like to use huge introductions and others, like me, are satisfied with small but noninvasive logos.

Before we start, you'll obviously need VirtualDub, and in this case for the watermark to work, the ffdshow codec. Yes, I'm afraid, yet another codec. You need this codec, because it provides filters we need, such as implanting the logo into a video. The ffdshow codec installs a plugin called "ffvdub.vdf" into your VirtualDub plugin directory. This plugin allows the access to ffdshows filters via VirtualDub.

VirtualDub has it's own internal logo filter, but the problem with this filter is it leaves a mess after encoding. The logo is barely readable, it's pixellated and all other sort of issues. Whereas the ffdshow filter leave a clear, readable logo instead.

Open your video, set your VirtualDub up to do whatever encoding your need. Here are two ideas. Now comes the logo part.

Adding a logo to a video

Open up Video -> Filters and add the filter "ffvdub" from Milan Cutka:

Now a new window will appear where all the available ffdshow filters are listed:

You're looking for the filter "Bitmap overlay". Enable it. Now select your *.bmp logo, place it to your liking via horizontal/vertical position and "Alignment". Experiment with "Mode" and "Opacity". Don't forget to enable preview it.

If you don't have a *.bmp logo or don't want to waste a lot of time, you can use the logo template I made a while back.

It's very important in what order the filters are listed! For example in the screenshot you see "Resize & aspect" followed by "Sharpen" followed by "Bitmap overlay". Although Resize and Sharpen are deactivated, if they were all enabled then the video FIRST is resized, THEN sharpened, THEN the logo is applied in this order.

This way the logo remains unaffected by Resize and Sharpen and retains its original quality (but in the end will ultimately be affected on how you set up your compression settings).

Here's how my logo looks: http://www.youtube.com/watch?v=SFBM5q98mwc (in the bottom right corner)

Fraps & VirtualDub: How to save disk space and enjoy good quality, small file size HD video clips

Don Salva — Mon, 07 Sep 2009 14:06:10 +0000

The biggest problem most of us face while recording gameplay clips with Fraps is its disk space usage. Fraps records everything uncompressed, continuously (free version is limited to 30 sec.) and sequential. Meaning from start till stop the "whole" clip is split into single 4GB files (For FAT32 filesystem user convenience).

The length of a single 4 GB clip depends on the resolution and the in-game scene you record it in. For example, for me a 4 GB clip has about 90 sec. worth of gameplay in 1920x1200 resolution.

Because of this uncompressed real-time recording your frame-rates drops significantly while playing. There is NO way to keep up 40+fps with Fraps running unless you have one hell of a monster system.

Imagine what happens when you want to capture a 30+ min gameplay clip. Either you buy yourself a Terabyte HDD or you use a very simple trick: recording in half-size, upscaling and sharpening. Read the How-To guide!

This guide is intended for 16:10 and 16:9 resolutions. It should not be used for 4:3 resolutions. One cannot upscale a 4:3 resolution into a 16:9, i.e. true 720p or 1080p, without the end result either having black bars left and right or being over-stretched/-squeezed.

Recording in Half-screen/size

Before we start with VirtualDub transcoding, upscaling and sharpening, we first have to record our gameplay. We do not record our gameplay in full-screen/size mode, as usual, but in half-screen/size:

This means if you have a game running at 1920x1200 resolution, Fraps will record it at 960x600, or if you run it at 1280x1024, Fraps will record 640x512, etc. Which in turn translates into much more gameplay time in each 4 GB file.

Another example: While recording in 1920x1200, I only got about 90 sec. (1:30 min) per 4 GB file, recording in half the resolution (i.e. 960x600) resulted in about 360 sec. (6 min) per 4 GB file. Whole 270 sec. more!

Upscaling and Sharpening

We want HD Quality, so we're going to upscale and sharpen our recorded clip. Start up VirtualDub, open the uncompressed file, change in "Video" -> "Full Processing Mode" and "Audio" -> "Full Processing Mode", then set your VirtualDub according to this screenshot:

Save your settings via "File" -> "Save processing settings" or Ctrl+S, so you can simply load them in the future. Saves a lot of time!

Addendum: Regarding "Threads"; QuadCores = 6, DualCores = 3.

You'll need VirtualDub and the H.264 codec, which you can get here. Get the VFW-version, and pay attention if you're downloading x86(32bit) or x86_64(64bit).

The settings in the screenshot are the same ones like in "VirtualDub 720p HD settings for YouTube! [Good quality, small file size]", what differs, however, is that we are also going to use the "sharpen" filter. If you want a very detailed explanation what each option in the x264 codec does, check this out!

We call the method to resize a lower resolution (960x600) to a higher resolution (1280x720, in this case), UPSCALING (sometimes also refereed to as up-converting). For your information, 1280x720 is the Standard HD (720p) resolution. The other is Ultra-HD or Full-HD (1080p, 1920x1080).

However due the upscaling method we lose quality, which is where the filter "sharpen" comes in. Upscaling from lower to higher resolutions causes the image to "blur" (A more detailed explanation). Sharpen counteract this.

You'll have to experiment with the slider in preview mode around a bit, until you find the result best to your liking. I found myself settling down between 23-30, depending on each clip.

When everything is set, start transcoding via "File" -> "Save as AVI".

Now on the other hand, if you UPSCALE from a lower resolution to HD resolution the video is not really HD material anymore, because you lose quality during upscaling. Just saying. It's at best pseudo HD.

Two important VirtualDub settings you should not forget:

"Options" -> "Preferences" -> "Threading": Set the "Video Compression Threads" to 1. This is important if you have a multi-core CPU to make use of the multiple cores and in turn speed up the transcoding process.

Since VirtualDub Version 1.9.4 there's an option to use your GPU as part of the transcoding process. Make use of it to additionally speed up your transcoding process. "Options" -> "Preferences" -> "3D accel", hack in the box.

Addendum 16.05.2010:
I've been getting mixed messages. Apparently using "3d accel" causes it to render slower than usual on some systems. I can confirm that. With 3d accel turned on, a 10min video takes about 1 hour 10-20 minutes, while turned off it takes about 1 hour max. You'll have to experiment on your system!

Addendum 23.10.2009:
There’s currently a problem causing VirtualDub 1.9.4 - 1.9.9, crashing under Windows 7. The culprit is 3D accel! Turn it off under “Preferences” -> “3D accel” and enjoy VirtualDub once again!

Update 27.01.2011:
Updated Setting screenshot

Real-Time Screen capturing/recording easily done with VirtualDub and VHScrCap

Don Salva — Mon, 10 Aug 2009 11:24:01 +0000

I've been looking for a way to capture my screen (not screen-shot it, but real-time capture) while on my OS (WinXP). I've found a lot solutions, but all of them involved shelling out money one way or another. Now why would I do that, when in the age of Open Source and Freeware there are free solutions for almost everything?

So I set out to look for an Open Source/Freeware solution, and I've found one! VirtualDub coupled with VH Screen Capture Driver works like a charm!

First you need to download and install VH Screen Capture Driver, then you start up VirtualDub.

In VirtualDub you do the following:

Step 1: Switch to Capture Mode

Step 2: Select Capture Device

This is why you need to install VH Screen Capture Driver, before VirtualDub, otherwise you won't have the Screen Capture "Driver".

Step 3: Set your capture file

Choose its name and its location.

Step 4: Configure the capture devide / filter

You need to config VH Screen Capture Driver in order for it to run correctly.

Press "select" then put in your current desktop resolution in Width / Height. I'm running my PC in 1920 x 1200 resolution.

And that's pretty much it. Didn't cost a cent!

You can record uncompressed audio/video and transcode it later:

or you can choose to expirement with those setting to your best liking! I personally record uncompressed and transcode it afterwards.

Enjoy!

Edit: According to some users, this method does not work with Windows 7 and Windows Vista.

Addendum 18.08.2010:
If you are a Windows 7 user, I suggest you use Fraps to record your desktop. Fraps has the ability to record your Aero desktop, which is awesome. No more hassle. Start Fraps, and start recording.

VirtualDub 720p HD settings for YouTube! [Good quality, small file size]

Don Salva — Sun, 01 Mar 2009 15:45:15 +0000

I often record gameplay clips from video games I'm playing with FRAPS. Usually in 1920x1080 resolution depending on the game and the FPS I get with FRAPS running. For transcoding the raw footage to something bearable to upload I usually use only VirtualDub.

Lately I've experimented with settings, and I think I found some good ones. Good video and sound quality with small file size were my goals. I think roughly 100 mb for a 5 min. clip and about 300mb for the full 10 min. in full 720p HD quality is awesome.

I'm still experimenting with sound compression but I think with 192 kb/s AC-3 ACM I have a good start.

Please bear in mind that the higher your resolution is the bigger your file size will be and vice versa. Those 100 mb respectively 300 mb are based on my 1920x1080 resolution!

This guide is intended for 16:10 and 16:9 resolutions. If you're playing on a 4:3 resolution, do not resize down to 1280*720p without maintaining aspect ratio, because the end-result will otherwise look over-stretched/-squeezed.

My settings for VirtualDub are in the picture below. If you want a very detailed explanation what each option does, check this out!

By the way, you have to have the H.264 codec installed to get it working. Download and install the latest VFW-version build here. PAY attention which version you're downloading, x86 (=32bit) or x86_64 (=64bit).

A warning about the Option "Multithreading" -> "Threads" don't set it higher than [# of your CPU Cores * 1.5] because then your whole system will start chugging. For example a Core 2 Duo would be 2 Cores * 1.5 = 3, a Q9550 would be 4 Cores * 1.5 = 6 and an AMD Phenom II X3 would be 3 Cores * 1.5 = 4.5 4.

This is by no means a guarantee that it will or will not work on your system. Each system, its components and its software are unique. You'll probably have to experiment a bit on your own. But those mentioned numbers are a good start.

It also might be necessary for you to adjust the audio volume to your liking. I have -10 dB, because the sound I record is terribad. My onboard sound sucks, period. Some people also experience asynchronous sound, here's how you can fix it.

On a side note: When you have uploaded your clip on YouTube and YouTube processed it the quality then will be lowered again. There's nothing you or I can do about that, it's up to YouTube.

The whole tutorial is also available on YouTube as a video:
http://www.youtube.com/watch?v=o9XOgnaJ_Hs

A few links to my own HD clips:
http://www.youtube.com/watch?v=nxijbXfIs4A&hd=1
http://www.youtube.com/watch?v=3yt35f2WFFY&hd=1
http://www.youtube.com/watch?v=Oe4jCQ8zzz0&hd=1
http://www.youtube.com/watch?v=eNLlobJbrm8&hd=1

Addendum 16.September 2009: Enabling GPU acceleration
Since version 1.9.4 it is possible to use the GPUs processing power (if your GPU supports it, which most modern GPUs do) as part of whole the transcoding process. This way you can speed up the transcoding process significantly (depending on your compression settings!).

On my computer this cut my transcoding time from an average 1h for a 10 min. clip down to some 40-ish minutes. Under "Options" -> "Preferences" enable "3D accel".

Addendum 23.10.2009: Crashes during encoding
There's currently a problem causing VirtualDub 1.9.4/5/6/7/8/9 to crash under Windows 7 while encoding sometimes. The culprit is 3D accel! Turn it off under "Preferences" -> "3D accel" and enjoy VirtualDub once again!

Addendum 16.05.2010: Option "3d accel" slowing down encoding
I've been getting mixed messages. Apparently using "3d accel" causes it to render slower than usual on some systems. I can confirm that. With 3d accel turned on, a 10min video takes about 1 hour 10-20 minutes, while turned off it takes about 1 hour max. You'll have to experiment on your system!

And if you are running on a multicore cpu, don't forget to enable threading via "Option" -> "Preferences" -> set "Threading" to 1!

Update 27.01.2011:
Updated Setting Screenshot

The Don's Craptastic World » virtualdub

Intel Core i7-2600K and encoding with VirtualDub

Related posts:

VirtualDub: How to use mkv, mp4, flv, mov and other file formats with VirtualDub

Related posts:

VirtualDub: Use ffdshow filters (resize, sharpen, wartermark) in VirtualDub

Related posts:

VirtualDub: How to prevent asynchronous sound

Related posts:

VirtualDub: How to split and merge video files

Related posts:

VirtualDub: How to watermark your videos

Related posts:

Fraps & VirtualDub: How to save disk space and enjoy good quality, small file size HD video clips

Related posts:

Real-Time Screen capturing/recording easily done with VirtualDub and VHScrCap

Related posts:

VirtualDub 720p HD settings for YouTube! [Good quality, small file size]

Related posts: