Jump to

Post Processing

Capturing the video is just the first step in getting the final encoded file. What lies ahead is the processing required to get the video ready for presentation.

Loading the Video:
Load the video segments in VirtualDub (loading the first file is usually enough as VirtualDub knows to automatically load additional segments).

Frame Rate Correction:
Some capture programs have a problem adjusting the video and audio clocks to match, resulting in frame rate inaccuracies.

To check if your capture was frame rate accurate, go to the Video -> Frame Rate dialog. If the Frame rate indicated by the "Change so video and audio durations match" setting doesn't match your selected frame rate, then enable this setting to adjust the frame rate to match the audio length.

If you don't enable this feature when the frame rate listed doesn't match your initial capture rate, the audio will go out of sync toward the end of the video.

Video Pruning:
With most video capture there is usually some junk frames you want to cut out. This is quite simple to do in VirtualDub, simply seek to the first frame in which you want to begin the cut, press the "Mark in" Button (looks like a small triangle with a line protruding from it), go to the last frame you want to cut, press the "Mark out" button (looks like a mirror of the "Mark in" button).

You should see part of the video time-line highlighted, then simply press the delete button on the keyboard and everything highlighted is removed from the video stream.

Video Post Processing:
Open the Video -> Filters dialog.

1: Field Swapping
If you notice that your video seems to have every second line reverse (and I don't mean the comb effect you get with Interlacing), then add a field swap filter.

2: DeInterlacing
DeInterlacing is the process which removes the Comb-Like effect you see when an object is moving or the camera pans. It is caused due to the differences in the image-display system used on a TV compared to a Computer Monitor.

There are several methods of cleaning the interlace effect. Some are faster than others, while some are only suited for a specific visual medium. It's also important to remember that DeInterlacing a PAL source can be different from DeInterlacing an NTSC source! It may also be possible that certain PAL content will not require DeInterlacing at all.

When dealing with a PAL source you'll want to first try using a PAL Phase-Shifting DeInterlace filter (one comes built into Donald Graft's Smart DeInterlace filter). Enable "Phase Shift" and make sure that "Disable motion processing" is checked. Try frame advancing a few frames. If the interlace is gone, good. If it isn't, try stepping a few frames backwards, if it's gone, it means you need to swap the field order before the DeInterlace filter, there's an option to do that within the filter.

If this doesn't work you can proceed to the Brute-Force method (also used for NTSC). The quality isn't as good, but it's as good as it's going to get. Disable "Phase Shift", make sure "Disable motion processing" is not checked and make sure "Blend instead of Interpolate" is checked. This will make the filter try to blend the interlaced areas of the screen. You'll see some ghosting when there's fast movement, but it's not really noticeable while playing at full speed.

Beside the Brute-Force method, with NTSC there's a process called IVTC which can rebuild an Interlaced 29.976fps video into a progressive (non-interlaced) 23.976 source. This gives the best quality. The process is quite complex, thus it has it's own tutorial, Video Capture 102 - IVTC.

Beside the above there is another issue. If you blend the video, it creates ghosting, which isn't really noticeable on real-life video, but it can be on cartoons. If you're encoding a cartoon, it's sometimes wiser to use VirtualDub's built-in deinterlacer with the option set to drop one of the fields.

3: Cropping
Ok, and here's the tricky part. A lot of the time when you capture video, the capture area contains unused area. It may be a few black pixels around the visible image, or even some random junk. This is part of the picture that serves no purpose and you should crop it out (and no I'm not referring to widescreen bars).

Highlight the first filter in the list and press the "Cropping" button on the right.

Now seek the video to an area within the show (the actual show you want to encode, not some commercial) which is bright enough so you can see the borders of the video area. And using the X+Y cropping options crop these unused areas out of the image.

4: Resizing
Two things that are extremely important about resizing. The first is that the image width needs to be a multiple of 32 (otherwise you get a massive CPU speed hit on playback [note: height doesn't need to be multiple of 32]) and the second is that the aspect ratio must fit the monitor, which means a 4:3 aspect ratio (Width/Height = 1.333...Infinity).

Add a Resize filter. Make sure the resize filter is set to the correct width and aspect ratio and that the filter mode is set to "Precise Bicubic" (best quality). I recommend setting the resolution to one of the following : [640x480, 608x456, 572x432, 512x384, 480x360, 384x288, 352x264].

5: Widescreen Cropping
Some video content is displayed as widescreen (black bars below and above the video area). At this point in the filter setup you can crop these out. What you do is add a "Null transform" filter. This filter doesn't really do anything, but gives us access to the cropping button AFTER the resize. If you're planning to add a logo then you can simply use the Cropping button on the logo filter instead. Once added, press the Cropping button and crop out the top and bottom black areas. Make sure not to alter the width! Due to the VirtualDub filter queue, it may be possible that the image that appears in the cropping dialog would not be the real image you are cropping, so you may want to double check the setting against the post-processed image in the right-video pane.

6: Logo
You can download a logo filter from the VirtualDub home page.

When doing a logo it's important to remember 2 things. Make it small, and make it short. Don't make it over 5-10 seconds in length and put it in a place where it won't annoy people (intense studies indicated that the lower-right corner is the least annoying). I put my logos during opening credits, but some people fear others will cut those out and claim it as their own cap, so make it small and in the corner if you're concerned.