Determine if changing encode option will be beneficial

"How to's" and tips and tricks - for all VSO products- not for questions!

Moderators: Maggie, Cougar_II, ckhouston, JJ, Forum admin

Post Reply
ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Determine if changing encode option will be beneficial

Post by ckhouston »

The Automatic encode option chooses from three encoding profiles, SP, MP or LP, the one chosen will show in the ConvertX work area and it is also shown in the log of the conversion. Automatic usually picks the best one except for videos that have very simple scenery. A method is presented here that will help anticipate the effects of manually changing to another profile.

The procedure relies on the attached plot. The plot was obtained by converting many different typical videos, covering the widely varying degrees of scene complexity that a user is likely to encounter, with all three encoding profiles, and using a very large custom target size to guarantee the conversion would not have to use extra compression to fit to a restricted size. There was some data scatter as expected so the curves in the plot were smoothed to fit the data in order to keep things simple.

Data posted by a user will be used to illustrate the most usual use of the plot. The user converted a 100 minute movie with the Automatic option and a DVD5 target size of 4300 MB. The Automatic option chose the MP profile because of the movie length and target size. That converted to a 2857 MB size which gives an average bitrate for the video of 2857 / 100 = 28.57 MB/minute. Finding that value on the horizontal axis and moving up to the intersection of the red curve and then left to the vertical axis gives an estimate of about 59.7 MB/minute for a conversion using SP and a non-restrictive large target size. That translates to about 59.7 x 100 = 5970 MB converted size. The user actually converted the same video with SP and a target of 10000 and got a converted size of 5765 MB, which is reasonably close to the estimated value and well within the scatter of data used for the plot.

Now the question is which profile/option will give the best quality. The best in this case would be to use a DVD9 7700 MB target. The Automatic option would then choose the SP profile resulting in the DVD being filled about 75% and give the best possible quality. But suppose you want to use a DVD5 4300 MB disc instead but still use SP by setting it manually. Then ConvertX would have to compress the unrestricted 5765 MB down to 4300, a reduction of 100 x 4300 / 5765 = 75%.

Compressing to about 80% usually does not cause a significant quality degradation. Two pass conversion should probably be used for anything less than 80% though, and I personally won't go below 65% because the more complex scenes in the video, if there are any, will suffer while the simple ones remain good. Going below 65% might work for videos with fairly low average bitrate, consisting of mostly simple scenes in other words like many TV shows for example, but you will have to try it and decide for yourself.

If you find that manually overriding Automatic improves the quality of a particular video, do not make the mistake of assuming that the same should be done for all videos, unless you always convert the same type of video, low complexity TV shows for example.

Notes:

1. Low bitrate in the plot means simple scenery, higher bitrate means more complex scenery. The video used in the example has about average complexity for example, its converted bitrate with MP is in the middle of the MP range. As shown in the plots, the average converted bitrate range for typical video is about 12 to 36 MB/minute with the LP profile, 16 to 46 with MP and 45 to 63 with SP.

2. The plot is based on data from conversions that did not fill the target, so any conversion you use with them must also not fill the target.

3. The plot can also be used in reverse of the example above. Suppose you did a conversion with a big target or with DVD9 which didn't fill the target and the Automatic option picked the SP profile. You could determine its average bitrate, find that value on the vertical axis, move horizontally to the intersection with the red curve, then down to the horizontal axis, and then multiply that value by the length of the video to find the size expected for a MP conversion.

4. There are some other ways of doing this in the Alternate Ways to Pick the Best Option section of http://forums.vso-software.fr/understan ... 13008.html. There is also a Judging Scene Complexity section there that may interest you if you are surprised by the level of scene complexity indicated for your video by these plots.
Attachments
Estimated bitrate vs Converted bitrate.png

Post Reply