Virtualdub directstream copies' bitrate often different from original file?

[egaddd]

Hello, this is my first post here! My question concerns using Virtualdub's directstream setting to cut up a video so that there is no loss in quality, i.e setting both video and audio to direct stream copy in Virtualdub.

I've often noticed that often a direct stream copy will have a bitrate that is somewhat smaller than the bitrate of the original. The difference between the new file and original varies depending on the individual file being copied (I've also created directstream copies that came out with a larger-sized bitrate than the original file).

Sometimes, however, the difference between the original and copy is more pronounced. At first, I assumed that the difference in bitrate (video and audio stream) was related to the rendering engine and/or processes used in Virtualdub to efficiently create the directstream copy (please forgive, my understanding of rendering and editing is still pretty rudimentary).

But now I'm concerned that some my directstream copies are losing quality. This is something I've noticed using several different versions of Virtualdub, and on both a Windows XP and Vista computer. For example, I just made a directstream copy using Virtualdub 1.8.8. The original file has a videostream bitrate of 1704kbps with a 33% overhead). The directstream copy has a bitrate of 1467kbps (with a 39%) overhead. And the difference in the appearance of the video of the directstream copy when compared to the original, to my eyes, is noticeable.

What's the source of the differences in bitrate when comparing directstream copies to the original, and why are some direct stream copies smaller than, some the same size, and some larger than the original file?

Thanx, I'm trying to get a better understanding of this and other related subjects. I appreciate any input.

-Egaddd

[jagabo]

But now I'm concerned that some my directstream copies are losing quality. This is something I've noticed using several different versions of Virtualdub, and on both a Windows XP and Vista computer. For example, I just made a directstream copy using Virtualdub 1.8.8. The original file has a videostream bitrate of 1704kbps with a 33% overhead). The directstream copy has a bitrate of 1467kbps (with a 39%) overhead.

Where are you getting those numbers? Those overhead values are enormous. They should be in the 1 percent range. Try using GSpot. It scans the video stream and reports a more accurate bitrate than most other programs.

And the difference in the appearance of the video of the directstream copy when compared to the original, to my eyes, is noticeable.

In Direct Stream Copy mode the compressed video frames are simply copied from the source file to the output file. They are no decompressed and recompressed. There is no change in quality.

How are you checking the quality? View both files at the same time in two media players? You will usually see obvious differences (even when viewing the same file in both players) because the first player will get the graphics card's video overlay feature, the second will be writing directly to the Desktop (only one player can use video overlay at a time). Video overlay has its own proc amp settings and may include other video processing features that can make the video look very different.

What's the source of the differences in bitrate when comparing directstream copies to the original, and why are some direct stream copies smaller than, some the same size, and some larger than the original file?

Many file info programs report the bitrate in the header of the video data, not the real average bitrate. In a VBR encoded file the bitrate in the header may be any random value. Use GSpot or look at the size of the files. Unless there is some other stream that VirtualDub isn't copying the sizes should be very close.

There are some AVI/ODML features that result in different amounts of overhead. Whether the program includes an ODML header or not, the size of the junk chunck to a align the first video frame on a 512, 1024, or 4096 byte boundary, the audio chunk size, etc. But these only change the overhead by small amounts (unless you have a really tiny file).

If you really want to get your hands dirty, VirtualDub has a RIFF Chunk Tree viewer (part of the Hex editor, under Tools) where you can examine all the headers and data chunks.

[egaddd]

Thanks for the explanation.

I got that overhead info and bitrate from checking the 'file info' in Virtualdub. Let me check the files, original and stream copy, in Gspot...ok, the overhead of the original in Gspot is .74% and the streamed copy is .85%. That's more in line with the sort of numbers you were expecting. Audio bitrate is 138kbps for both in Gspot, video bitrate is 1704kbps and 1466kbps respectively.

As far as checking the quality?! I opened one, closed it, then opened the other, over and over, at the same portion of the film, and eyeballed it. Very scientific and unbiased, no doubt! But I grok what you're saying. Still, shouldn't the video bitrates also be the same, or close to the same?

Edit: I just re-read your post, and see that you addressed this. Ok, good info.

[jagabo]

You can use an AviSynth script like this to compare two videos:

v1 = AviSource("video1.avi")
v2 = AviSource("video2.avi")
sub = v1.subtract(v2) 
substrong = sub.levels(112,1,144,0,255) 

return (StackVertical(StackHorizontal(v1.subtitle("original"),v2.subtitle("encoded")),StackHorizontal(sub.subtitle("Difference"),substrong.subtitle("Difference amplified"))))