Method 1: Creating Clips for Source Footage:

I'm sure you've already read the codec guide, if not do so now as it's a handy general overview of different video codecs.

Before I actually get onto the conversion details I would like to emphasis a few things:

DO NOT USE DIVX or any divx variant. They are very difficult to edit with and the quality, no matter how good it looks to you when you are editing, will make things look worse than if you used a lossless compression codec. MPEG4 codecs may well be small but they are not good things to edit with.

DO NOT USE MPEG1 - lots of people like using mpeg1 for editing. However, even in Premiere and other programs that have mpeg1 support, the editing will not be frame accurate and you may find that the cut you thought was spot-on actually pops into the next scene when editing. This is very bad, not to mention that the quality will be pretty awful when it's all done too.

Also, do not use mjpeg unless you really have to - if you have the space for lossless compression then please use a lossless codec. If you MUST use a lossy codec like MJPEG, choose a very high quality. If you want to make DV files, use your native software/hardware solution. If you want to use anything else, for gods sake at least make sure all the frames are keyframes :)

Preparing the Clips

What you should have:

If you have followed the guide so far you should have an avisynth script that

• Imports Footage
• (optional but recommended) Removes Interlacing (makes the footage progressive)
  
• Corrects any Aspect Ratio issues
• Is generally nice, clean and looking good
  

Converting the Avisynth File

At this point, you should have an AVIsynth file (that is either progressive or interlaced). Load up VirtualDubMod and open the avisynth file.

Choosing a codec

Right now you essentially have two choices. HuffYUV or Lagarith.

OK I tell a lie, there are a number of options available to you but HuffYUV is without a doubt the most stable of these options and Lagarith is the second most stable. However, I will talk about some alternatives.

The main problem with HuffYUV as it is at the moment is that it does not have the ability to store YV12 data. YV12 is the native colorspace of DVD and is hence the format that would be most appropriate to store the colour information in. It also contains physically less data due to the fewer chroma samples per frame.

Unfortunately we cannot take advantage of this with HuffYUV as it does not support YV12. This is a major failing that I would like to see fixed. However, in its wake there are a number of lossless compression codecs that can  store YV12 information.

VBLE - This is a codec by MarcFD but it is unstable. It works but has lots of bugs (particularly with directshow) and as there is no source code for these bugs to be fixed unless MarcFD releases newer versions himself.

HuffYYV YV12 - That's right, there is a version of HuffYUV that can compress YV12 data. It is part of the VfW code for FFDShow and be selected amongst the massive amounts of options in the FFDShow codec configuration pages accessible via VirtualDubMod. The only problem is that  FFDShow is under heavy development so it is possible that you will install a version in the future that will break compatibility with old encodes. If a stand alone version of this codec can be produced then I will recommend it as the code itself is sound.

ffv1 - This is another codec in FFDShow and very powerful it is too. This is the best YV12 video compressor there is at the moment but it is very developmental and undergoes constant changes. As such, my concerns are the same as with the FFDShow HuffYUV.

CorePNG - A really great codec for anime. However, it has not been updated in around 5 months and has some known bugs. Even with this resolved, however, it is very slow both encoding and decoding so it's not vastly useful for editing.

Loco - Again buggy and this time also slow.

Lagarith Lossless Video Codec - Good compression and becoming more stable all the time.

DVD clips are YV12 and you will have the smallest encodes with YV12 compression. For this you will probably want to use Lagarith as it is the most stable of the YV12-capable compressors available. However, it is not the fastest of codecs so if speed is your primary concern then you may wish to convert to RGB and compress with HuffYUV. 

Both of these codecs come as standard with the AMVapp.

Compressing with Lagarith in YV12 (smaller files)

Open up VirtualDubMod and in the menus go to  Video -> Compressor and in the list choose Lagarith. The options are very simple:

Enable Null Frames: This option will increase compression but the files it produces can be problematic to edit with, so it is not recommended for stability reasons.

Enable RGBA Compression: This isn't required for YV12 encoding but selecting this will allow you to also store the alpha channel when compressing RGB32 video (from After Effects etc).

Always Suggest RGB Format for Output: This is a decoding option which will feed RGB footage to programs instead of the native colourspace that is stored. This is to accomodate for programs that do not like YV12 or YUY2 and hopefully you should not need this.

So, for most people, none of these options will need selecting.

Once you have the compressor selected you must make sure that you select Video -> Fast Recompress which will ensure that there will be no colourspace conversions between your input (from Avisynth) to your output (Lagarith). Lagarith can compress YV12, YUY2 and RGB but if you are using DVD footage then (unless you have changed the colourspace in avisynth somehow) you will compress in YV12 which will be identical to the dvd source.

For making the clips, see the section below...

Compressing with HuffYUV in RGB (faster files)

As we will not be using a YV12 codec, a colorspace conversion must take place when we compress DVD MPEG2 clips into HuffYUV. HuffYUV can store YUY2 and RGB colour. RGB is the colorspace used by the majority of video editing software out there - the advantage of saving clips in RGB is that you can be certain that what you put into the video editor is what you will get out. YUY2, however, is naturally smaller than RGB and will produce smaller clips which is good if you do not have much room.

Setting up HuffYUV

Open up VirtualDubMod and in the menus go to  Video -> Compressor and in the list choose HuffYUV. 2.1.1 is the current recommended version as 2.2.0 is buggy. If you have 4 versions of HuffYUV displayed, then you have the DirectX 8.1 Yuv Renderer Fix installed. Don't worry, they are just separate listings of the same codec. Once HuffYUV is highlighted, click the Configuration button. HuffYUV has several options which I will explain below.

YUY2 Compression Method: HuffYUV can compress YUY2 colour. Although DVDs use YV12 colour, you can convert to YUY2 losslessly provided you use the appropriate convertion command in avisynth (see below). Setting this to Predict Median (best) will allow it to use the highest compression when asked to compress YUY2 footage and it is smaller than RGB. RGB, however, is better for editing compatibility in many cases.

RGB Compression Method: You should always have this set to Predict Median (best). It is likely that you will mostly be compressing RGB footage with HuffYUV as that is the colourspace used by most editing software. Making clips in YUY2 is fine but if you export from an editing program you should always, without exception, make sure that the RGB mode has a compression method.

Field Threshold: This is the number of lines before HuffYUV will consider the file to be interlaced and compress fields instead of frames. Honestly it doesn't matter much either way except you will get smaller files if you do have interlaced footage. To always have HuffYUV always presume the frames are progressive, change this to a number higher than 576. 'Ignore is-interlaced flag' is a decoding option that can be helpful if a huffyuv file appears to be decoding incorrectly as this can sometimes fix it.

Always Suggest RGB mode for Output: Leave this unchecked unless an editing program is not displaying your encode (such as After Effects in some cases). It wont affect your encode at all, it's a decoding option.

Enable RGBA Compression: Alpha channels can be very useful things if you know what they are and how to use them. For footage conversion an Alpha Channel will just waste space but when rendering After Effects compositions or 3D renders, the Alpha Channel can be handy.  Selecting this compresses the footage as RGB32 instead of RGB24. However, if you leave it on all the time it is wasteful and unless you have an alpha channel it will make your encodes bigger than they need to be.
 
Swap Fields on Decompress: You should not need to use this - it is designed to fix problems with bad capture cards that have encoded with a field order opposite of what the source was.

Enable Console-window logging: Again, you won't need this.

YUY2 or RGB

At this point you need to decide whether you want maximum compatibility or you want better file sizes. If you compress RGB with huffyuv you can be certain that what you put into a video editor is what you will get out. With YUY2, this is true most of the time but some programs can do bad conversions or simply refuse the YUY2 footage. You can often get around this downfall with the "Always suggest RGB format for output" but it's not as reliable as having clips in RGB. So, essentailly, it comes down to space. If you have space to spare, use RGB but if you are tight on space use YUY2.

Converting the Colourspace - different commands for different sources.

If you have used IVTC or Force FILM on your footage, then setting up your avisynth script for compressing is easy. At the end of your script you should add a line that converts the footage to the colourspace you will be compressing to.

So for YUY2 use ConvertToYUY2() and for RGB use ConvertToRGB24()

This should be the final line of your avisynth script.

For progrsessive (deinterlaced/IVTC'd) footage:

ConvertToRGB24() or ConvertToYUY2()

For completely interlaced footage:

ConvertToRGB24(interlaced=true) or ConvertToYUY2(interlaced=true)

For Hybrid footage (these are AMVapp-only commands):

ConvertHybridToRGB24() or ConvertHybridToYUY2()

Once you have this footage loaded into VirtualDubMod, you should Choose Video > Fast Recompress from the menu and choose the Compressor... as HuffYUV and make sure it is set up as displayed in the image above. You should now be ready to make your clips.

Making the Clips

Once you've got the codec and the compression method set up, you can search through your source using the bar at the bottom and then use the two rightmost buttons (Mark In and Mark Out) to select the area you want to br converted into a clip. After you have chosen the range, simply go into the File menu and choose "Save as avi..."

You may remember in the VirtualDubMod guide that I talked about the Job Control. Well job control is definitely something you will want to use here. Instead of saving to disc, add them all your your job control and then run them in one batch later. It's much quiker than waiting for each clip to encode.

You will notice that these files will be very large - much much larger than the DVD source you are taking them from. This is because DVDs use a lossy compression method, whereas codecs such as Lagarith and HuffYUV are keeping every bit of information without losing any quality. To store video like this you need a lot of space.

Repeat as necessary.

Note: If you are running Windows 9x or have a FAT32 formatted hard drive, then you cannot create clips that are larger than 2 gigabyte (as they can't be read by that drive format). Either make smaller clips or use virtualdub's "Save as Segmented Avi" option which will do it automatically for you.

Now that you're done, you should have a look at how you set up a project in your video editing application.

AbsoluteDestiny - Nov 2004