VirtualDub questions & answers

[zorin]

To understand VirtualDub's behavior when cutting in direct stream mode requires a bit of knowledge about the structure of compressed digital video.
Simply put compressed digital video consists of key frames and (to keep matters really simple) non-key frames.
In a key frame is recorded the entire video picture at a certain point in time. In the non-key frames that follow that key frame only the changes in the video image over time with respect to that key frame are recorded. So a non-key frame never contains the complete video picture.

In direct stream mode VirtualDub simply makes a 1:1 copy of all frames between the start and end point you've set for the cut you want to make.
For that to result in a playable video clip it needs to start at a key frame otherwise when viewing you'd end up with garbage on your screen until the first copied key frame comes along.
If your video clip were to start at a non-key frame your video player wouldn't know how to construct the picture in the viewer because the non-key frame describes changes to a key frame that is not in the video clip.

In full processing mode the resulting video clip is constructed by re-encoding through which a completely new key frame is constructed at the exact point of your cut (where previously there was none).

Hope the above makes matters a bit clearer.

[prof_derzshowitz]

Originally Posted by zorin

... compressed digital video consists of keyframes and ... non-keyframes.

[Recorded in] a keyframe is... the entire video picture at a certain point in time. In the non-keyframes that follow that keyframe[,] only the changes in the video image over time with respect to that keyframe are recorded. So a non-keyframe never contains the complete video picture.

In direct stream mode [emphasis added by P.D.] VirtualDub simply makes a 1:1 copy of all frames between the start and end point you've set for the cut you want to make. For that to result in a playable video clip it needs to start at a keyframe...

In full processing mode [emphasis added by P.D.] the resulting video clip is constructed by re-encoding through which a completely new keyframe is constructed at the exact point of your cut (where previously there was none).

Thank you, Zorin, that does help to explain things quite a bit. So, if I have understood things correctly, by compiling the information included in your post, and in the earlier contributions of Shirtfull, the following is (basically) true:

1 - Compressed digital videos (such as .avi's, .mpeg's, etc.) consist of Key-Frames (which include the complete video picture at a given moment) and Non Key-Frames (the remaining image frames that change with respect to the preceeding Key-Frame).


2 - When using Direct Stream Copy to render a segment that you have marked with start and end points, an existing Key-Frame, that is already present in the video file, will be selected as the initial playing frame, since DSC involves a 1:1 copy of the original frames in the file.

RESULT: You end up with a significantly smaller video file (GOOD), but your selected start and end points may not end up where you had wanted them (BAD).


3 - In contrast, when you render a particular segment using Full Processing Mode, new Key-Frames are created, (most importantly) including a new KF at the location that you have marked as the start point for the new video.

RESULT: Your exported video file starts at the proper frame (GOOD), but will end up being unnecessarily large (BAD).


4 - The Smart Rendering function combines the two modes, resulting in a process wherein a new Key-Frame will be created at your selected starting point on the timeline (through the selective use of FPM), while the rest of the video (apart from the initial re-encoded sections) will be copied using DSC to limit file size.

RESULT: Your new video starts where it should (GREAT), without resulting in an overly large video file (GREAT).

---

Alright, I think that covers everything. If I've misunderstood something, please let me know. Otherwise, I guess it's time to start editing some movies!

Thanks a lot, everybody, for the helpful information. It's pretty incredible how much you can learn about software and technology when dealing with pornography.

[2Bizzy]

Originally Posted by prof_derzshowitz

4 - The Smart Rendering function combines the two modes, resulting in a process wherein a new Key-Frame will be created at your selected starting point on the timeline (through the selective use of FPM), while the rest of the video (apart from the initial re-encoded sections) will be copied using DSC to limit file size.

RESULT: Your new video starts where it should (GREAT), without resulting in an overly large video file (GREAT).

That sounds like the best option then, ill go with that one, thank you very much guys I really appreciate it.

[zorin]

Originally Posted by prof_derzshowitz

Thank you, Zorin, that does help to explain things quite a bit.

My pleasure.

So, if I have understood things correctly, by compiling the information included in your post, and in the earlier contributions of Shirtfull, the following is (basically) true:

1 - Compressed digital videos (such as .avi's, .mpeg's, etc.) consist of Key-Frames (which include the complete video picture at a given moment) and Non Key-Frames (the remaining image frames that change with respect to the preceeding Key-Frame).

Non-key frames describe the things that have changed in the image with respect to the preceding key frame. They don't contain full image data.

E.g. a key frame holds the full image data of a house with a garden in front, a clear blue sky with a couple of birds (the flying kind), a cat asleep on the lawn and a man standing in front of the house.
The next non-key frame would simply contain : man raises right arm.
The non-key frame after that : right hand moves from left to right and back.
In the viewer of your video player you would see the man standing in front of the aforementioned house, raising his right arm and waving.

<snip>Direct Stream Copy <snip> involves a 1:1 copy of the original frames in the file.
RESULT: You end up with a significantly smaller video file (GOOD), but your selected start and end points may not end up where you had wanted them (BAD).

Actually only the start point is moved to the nearest keyframe when cutting in DSC mode. The end point can be on a non-key frame as this won't affect playback.

Full Processing Mode<snip>
RESULT: Your exported video file starts at the proper frame (GOOD), but will end up being unnecessarily large (BAD).

Only if you neglect to set the compression parameters via the video menu as this has a default setting of uncompressed RGB which would account for the large file.
Choosing a video codec and setting its parameters will result in a compressed avi file the size of which depends on the compression settings.

Haven't used smart render myself so can't comment on that.

[chiwa]

Originally Posted by zorin

Actually only the start point is moved to the nearest keyframe when cutting in DSC mode. The end point can be on a non-key frame as this won't affect playback.

That has been my experience clipping scenes with VDub. The start point has to be on a key frame, but the end can be moved. Nice to know that is the way program works, and I wasn't f'n it up more.

I have learned the little bit that I know here in the guide section, so thanks to everyone who helps us out.

[zorin]

Originally Posted by chiwa

That has been my experience clipping scenes with VDub. The start point has to be on a key frame, but the end can be moved. Nice to know that is the way program works, and I wasn't f'n it up more.

To be more accurate the above is true when the clip you to want to keep is marked by the start and end point you've set.

When marking a clip to be deleted from the parts you want to keep, the reverse is true.

E.g. if you have a video with 4 scenes and you want to cut out scene 2. The start point of the cut can be on a non-key frame somewhere in scene 1 but the end point of the cut needs to be on a key frame at the beginning of scene 3.

With the cut marked like this and pressing the delete key in VDub scene 2 will be deleted from the video which can then be saved in DSC mode.
The resulting video will playback without a problem because scene 1 ends on a non-key frame and is followed by (the previous) scene 3 which starts at a key frame.

[haldane4]

Thanks for this great thread - I was attempting some edits with virtualdub this week and the innaccuracy was driving me crazy. Hope I can get it sorted now.

[haldane4]

Has anyone used the Neat Video plugin for Virtualdub? It removes a great deal of picture noise from VHS transfers working frame by frame - the only problem is it takes like 6 hours to process a 30 minute clip. Just wondering if this is normal.

Wish I had a Quad core.

[zorin]

Originally Posted by haldane4

Has anyone used the Neat Video plugin for Virtualdub? It removes a great deal of picture noise from VHS transfers working frame by frame - the only problem is it takes like 6 hours to process a 30 minute clip. Just wondering if this is normal.

Assuming it is a PAL video (25 fps) that would amount to (30x60x25)/(6x60x60)=2.08 frames processed per second, which (depending on your processor speed) would seem reasonable to me.

Wish I had a Quad core.

That will only speed things up if the plugin is specifically designed to use more than one core when they're available. Don't know if that is the case.