Bitrate quality question

[The Old Timer]

What percentage of the original file size should I be expecting when recoding into MPEG4 for approximately similar quality? 25%, 20% etc?

Normal commercial DVD's always seem to be are encoded with a bit rate of less than 8000. The bit rates recorded by my PC digital TV tuner card are often up to 15000. Does starting with this much higher bit rate mean that I could set the MPEG4 encoder to make a file that is much lower percentage as opposed to a regular MPEG2 original? Confused?

[jggimi]

It's not the percentage of the original filesize that we typically use, since compressibility is content dependent. Instead, we use a percentage of the maximum possible filesize at a given resolution. This technique is called a "compressibility test." The metric used is bits / pixels * frames, which provides a bitrate densitiy for any content in a single value.

AutoGK does a compressibility test for you, and if I recall correctly, aims for 75% of maximum. With GK, its optional (though highly recommended), and while Doom9's GK guide recommends 40-60% values, many members prefer b/p*f values that are 60-80% of maximum.

[The Old Timer]

If I interpret what you said correctly then it doesn't matter what in input bit rate is it's more a case of the quality of the input that matters. Furthermore AGK is selecting output settings purely on what is perceived to be the good quality setting. The figure that the analysis comes up with is what? The percentage of the date in the original file that is being used? Correct.


When you use preview this is done on the basis of what the program will be using for for the final encoding based on the file size or percentage that you nominated in the first pace? If for instance you select to fit on 2 CD's & you do a preview & then choose a different setting of say 1/3 of a DVD is the first analysis still used or discarded?

[CWR03]

If I interpret what you said correctly then it doesn't matter what in input bit rate is it's more a case of the quality of the input that matters.
Input quality helps, but only to an extent. Final output quality is much more dependent on final bitrate and resolution, which in your case is determined by filesize specifications. Naturally if you change the output size after your first analysis, it's no longer valid.

[jggimi]

The figure that the analysis comes up with is what? The percentage of the date in the original file that is being used? Correct.
I'll try and be clearer, because it looks like I've confused you.

AGK and GK both use a single numeric value that describes bit density of the output video. The total number of bits in the video stream, divided by the pixels in each frame, times the number of frames. Bits / pixels * frames, or b/p*f.

Your content sets the number of frames; but you select the number of bits in the output -- bitrate = output size. You may also decide to have no control, partial control, or complete control over your resolution, depending on your use of AGK or GK.

A compressibility test will take a sampling from the entire content; typically 5% (by selecting a 14 frame sequence every 280 frames).

The test is run at an initial resolution, and produces log output that determines the maximum possible bitrate for that resolution with the codec -- the highest bit density possible.

With AutoGK, it will resize (changing one of the factors in the forumula) within the limits you allow, and/or change codec features / video filters, to try to produce a video that has a b/p*f value as close to it's ideal of 75% of the maximum obtained during the compressibility test. Yes, it will factor in your new resolution, the p in the b/p*f formula, all automatically. When making codec feature changes, if I recall correctly, it will estimate your new compressibility based on its knowledge of the codec feature changes it decides to make.

GK allows total control of the process. You select an intial resolution, run the compressibility test, then select changes in resolution or bitrate (file size) to reach your goal. If you wish to change codec settings or change video filters, you must re-run your compressibility test yourself.

[Sagittaire]

Well bpf method is not correct
http://forum.doom9.org/showthread.php?t=95122

[jggimi]

In reading responses to your theory in that thread, it seems that your opinion does not yet have a complete consensus.

In our case here, I'm referring to b/p*f as is used with AutoGK and GK. In that thread, you start your consideration of b/p*f problems by saying:...the source compressibility is not predictible...And that's true, which is why compressibility tests are run with AGK and GK. (Yes, they're optional with GK, but those who skip testing with GK quickly run into trouble. Using any b/p*f value without a test of the source is not best practice.)