Encoding Q and A

[ckhouston]

A lot of people come here believing that traditional VBR encodes are better than CQ encodes mostly because VBR is the method they are familiar with and VBR will completely fill their DVD while CQ does not -- see http://forums.vso-software.fr/understan ... 13008.html for an explanation of the differences between them. This thread may help them reconsider that belief.


1. Q: Which method does ConvertX use?

A: It uses both. A CQ encode is done for projects that won’t completely fill the target disc and a VBR encode is done for projects that will fill the disc. The Automatic encode option will choose an encoding profile that will result in a CQ encode when possible for typical projects, but will switch to a VBR encode when filling the target can’t be avoided.


2. Q: Can VBR be forced in ConvertX?

A: Yes for most projects except very short ones. Set the encode option to SP instead of Automatic and you should turn on Two pass encoding at the same time. This is advisable only for some projects though so do yourself a favor and read the rest of this topic to be sure that is what you really want to do.


3. Q: Doesn’t a two pass VBR encode give the best possible distribution of bitrate to each scene according to their complexity?

A: Absolutely not. A CQ encode is free to vary bitrate as needed to give the same quality for all scenes by holding a parameter called quantization factor constant. A VBR encode is constrained to convert to a specified size so its assignment of bitrate has to vary less by necessity than with CQ in order to fit the result to the specified size. That results in higher bitrate for simple scenes than with CQ but lower in complex scenes, resulting in high quality in simple scenes but lower quality in complex ones. A second pass provides a better assessment of how resources can be used and still fill the target, which allows a bigger swing in bitrate between scenes, which results in more consistent quality for all scenes. A third pass allows even bigger bitrate swings and consequently even more consistent quality. In fact, if enough passes are used, a very good VBR encoder will approach a constant quality bitrate distribution similar to that produced by a quick single pass CQ encode.

Users should be aware that forced VBR in ConvertX is not quite the same as traditional VBR in most other programs. The bitrate distribution with a single pass forced VBR is usually a little flatter than with traditional VBR, so a second pass is essential if you do force VBR.


4. Q: So which is better, a CQ encode that provides the same quality for all scenes or a VBR encode where the quality is higher for simple scenes but lower for the complex scenes?

A: CQ normally gives better overall quality for all scenes in typical video where scene complexity varies a lot. Overall quality of VBR encodes will be inferior for typical video where scene complexity varies, but will give the best result for video consisting mostly of simple scenes like TV episodes.


5. Q: How do I decide which encode option to use?

A: Average users should stick with the Automatic option, it will produce very good quality for typical projects -- the Two pass option is usually a waste of time in that case. But you can get slightly better quality by forcing VBR for projects containing mostly simple scenery. There is information in http://forums.vso-software.fr/understan ... 13008.html and in http://forums.vso-software.fr/determine ... 17528.html that will help you decide if the effort is worth it.

Note: It is a common misconception that action primarily determines scene complexity, it usually does not -- see the Judging Scene Complexity section in the first linked topic. The average complexity for the entire video is most important anyway and the plots in the second topic are useful for judging that.


6. Q: But isn’t a VBR encode that fills my DVD always better than a CQ encode that only fills half of it?

A: No, it is better only in limited cases as already discussed. But if you have carefully considered everything above including the separate threads that are referenced and still ask that question, then you will probably be more satisfied with VBR no matter what is said. Be aware though that your bias in believing that bigger must somehow always be better influences your opinion when you evaluate conversions. Other people just as expert as you, or even more expert, that don't have that bias might disagree in their evaluation of the same conversion you watch.

Judging a conversion only on size ignores the fact that CQ can be more efficient than VBR in distributing bitrate among different scenes. And the fact is that a VBR conversion, even a two pass one, favors simple scenes more than complex ones. It may assign more than four times as much bitrate to a simple scene than CQ will but provide very little or no noticeable improvement. Then bitrate must be drastically cut in complex scenes to compensate even though converted size is maybe twice as big as with CQ. So complex scenes may suffer drastically as a result.


7. Q: Then why are CQ encodes not used more often?

A: Probably because they are more difficult to do than VBR encodes. Most good encoders have a CQ mode that is seldom used because it requires more expertise and some trial and error to prevent the converted result from filling the target -- a CQ encode is not possible if it does fill the target. ConvertX is the only program I'm aware of that has a one click CQ encode that anyone can use regardless of their level of expertise.


8. Q: What method is used on commercial DVDs and why?

A: I have only examined three of them from major studios. It appears that they use VBR in order to fit all features they want to include on the DVD. But, and this is important, it appears that they used a very good VBR encoder and performed multiple passes because the quality, as indicated by the quantization factor, of the different scenes was nearly constant.


9. Q: What factors affect converted size?

A: Size is proportional to the total play time of your project, a project twice as long as another will produce double the size if all other factors are equal. More complex scenery gives bigger size than simple scenery, more than three times more is possible. The encoding option, SP, MP or LP actually used can make as much as 375 % difference under some conditions. And the choice of resize filter can make as much as 35% difference. The effect of other factors is much smaller than those four.


10. Q: Is 2 pass ever necessary in Automatic encode conversions?

A: Only for some very long projects as explained in the 2 pass section of answers-to-most-of-your-conversion-questions-t16457.html.


11. Q: Why is my converted result so much bigger (or smaller) than my source file?

A One big factor is a change in resolution. Changing from a low resolution in the source file to a higher one used on the DVD will tend to make the DVD bigger. Changing from a higher resolution to a lower one on the DVD will tend to make the DVD smaller.

Another factor is the efficiency of the encoding format. Most source files were probably encoded in formats that are more efficient than the MPEG-2 format used on DVDs. That tends to make DVDs bigger.

And another factor is the bitrates used to encode the source and the DVD. The person that encoded the source file could probably choose to encode in a high or low bitrate. And bitrate varies in ConvertX as explained in some of the answers above.

The net result is that comparing sizes of sources and DVDs is complicated and usually a waste of time. An image is attached that shows sizes of three files that have the same content but different resolutions, bitrate and file type. But all three files convert to about the same size of 8 MB. So the size of the DVD ranged from about 10 times bigger than the source to about 20 times smaller in those three cases.