High Definition AutoGK settings

[NMachiavelli]

For those who use AutoGK on HD material, what settings do you use?

I'm essentially a beginner when it comes to video compression, which is why I really appreciate the ease of use of AutoGK.

I archive to hard drive not to optical media, so I don't care about achieving a specific size. For DVD material I set Target Quality of 80% - 100%. When I use these settings on HD material, though, the resulting avi file is larger than the original ts file.

I'm reluctant to set a custom size, since not only will that require multiple passes, but I don't want to necessarily assume that a certain percentage of the original size is going to be optimal.

I'm assuming people that work with HD are going to be primarily interested in retaining high quality. What do you do? Ratchet down on the target quality setting to 50% or so, or downsize the resolution in the advanced settings and let the computer's scaler make up the difference on the HDTV display?

I've done a lot of searching and I haven't found much discussion of people using HD content and AutoGK. Maybe the better answer is to start learning and playing with AVC tools?

Thank you.

[laserfan]

The real issue as I see it--converting HD takes an incredible amount of time, whether converting to a >SD i.e. HD file or something that's playable on a standard def system.

I routinely convert HD transport streams of TV shows to SD Xvid files--takes over 4 hours for a 44min program. The filesize goes from 4.7Gb to 340Mb (!) and looks fine. I think the one HD-to-HD I tried took almost twelve hours--the filesize was about half, could probably have been much less, but the time involved is such that I'm inclined to just save the original .ts as-is (to DVD-5 or -9).

[nix0]

It depends on what you want as output - is maintaing the resolution important? I would assume keeping the quality percentage high should work - though when I've wanted to convert a HD source to a player friendly SD xvid I used Gordion Knot as I was assurred of getting the settings I wanted

[NMachiavelli]

It depends on what you want as output - is maintaing the resolution important? I would assume keeping the quality percentage high should work...
The point of using AutoGK is to easily and simply take advantage of the presumably better compression of mpeg4 instead of the standard mpeg2 of current HDTV files, which I would assume would result in a smaller file size with little to no loss in picture quality.

I've already stated that my tests show high quality percentage settings result in a larger files than the original file. That is beyond useless. I would think that this shouldn't happen regardless, and that the only reason it does is that AutoGK makes incorrect assumptions somewhere when it comes to HD files, but I really don't know.

Maintaining the resolution isn't as important as maintaining the picture quality. I've found that watching a DVD on my 17" 1900x1200 LCD is almost as nice as watching taped HD content because, as I understand it, the computer is able to upscale the DVD resolution so well. If the same applies to an HD file, then I theoretically would have no problem reducing the resolution and letting the computer "fill in the gaps."

I'm mostly curious if this is what other people have to do in order to get high quality mpeg4 from HD files. As it stands, it seems that AutoGK is simply the wrong tool for the job if I want to maintain HD level quality in my compressions.

Thank you for the responses.

[HookedOnTV]

The quality based encodig in AGK is really fairly useless. The results are going to be varied depending on the source material.

Absolute best method is 2 pass with settings determined from doing a compressability test. Fiddle with output resolution and bitrate to get the quality value to near 100% (yes some will say that near 100% is overkill but I really notice the difference) in GK. Problem with this method is it takes a lot of time to set up each encode as they are all different and the encoding time can be crazy.

After trying just about everything on over a terabyte of HD ts files, many done with the two pass GK method I have recently settled on doing single pass constant bitrate, 6Mbps, with an output resolution of 960x540 and this is producing good results in around 4 hours.

[len0x]

The point of using AutoGK is to easily and simply take advantage of the presumably better compression of mpeg4 instead of the standard mpeg2 of current HDTV files, which I would assume would result in a smaller file size with little to no loss in picture quality.


While it might seem reasonable - its totally incorrect assumption. If you have heavily compressed mpeg2 source to start with - its very probably that mpeg4 size will be much higher than original. Main factor is: mpeg2 and mpeg4 do not have the same quality/quantizer metric, so you can't compare them directly. Advantage of mpeg4 is that if you find two sizes(one for mpeg2 and another for mpeg4) that look visually the same, then mpeg4 sample will have smaller size, but this will not mean they would have the same "quality" metric. Ability of a codec preserve details have nothing to do with raising/lowering quality. It is found empirically that 75% is a good mpeg4 quality percentage for SD resolutions. It might not be true for HD material where you indeed may be in need of just 50% to still look good. You can judge that by your own eyes and find threshold that suits your needs (and yes by setting quality to 100% you will be waisting quite a bit of bitrate...)

[len0x]

The quality based encodig in AGK is really fairly useless. The results are going to be varied depending on the source material.


I like this statement :) What did you expect? It just proves that you need a comp test to judge the source - nothing more, nothing less. If you don't have space issues you can perfectly use it...

After trying just about everything on over a terabyte of HD ts files, many done with the two pass GK method I have recently settled on doing single pass constant bitrate, 6Mbps, with an output resolution of 960x540 and this is producing good results in around 4 hours.

So you can tell the difference between say 90 and 100% quality encode and yet you are very satisfied with CBR encoding? C'mon, we live in the 21st century - forget the word CBR :)

[NMachiavelli]

len0x, thank you for clarifying my misperceptions...

It is found empirically that 75% is a good mpeg4 quality percentage for SD resolutions. It might not be true for HD material where you indeed may be in need of just 50% to still look good. You can judge that by your own eyes and find threshold that suits your needs...
Yes, and this is what I was hoping to determine from others who use AutoGK on HD material - what they use, and what is sufficient for them. So far, I haven't heard from anyone who routinely uses AGK to archive HD material.

[HookedOnTV]

I like this statement :) What did you expect? It just proves that you need a comp test to judge the source - nothing more, nothing less. If you don't have space issues you can perfectly use it...

Certainly didn't mean any disrespect to your tool. It's very good and I use it. Just meant that single pass percentage encoding is unpredictable. If space weren't an issue there would be no reason to encode it.

So you can tell the difference between say 90 and 100% quality encode and yet you are very satisfied with CBR encoding? C'mon, we live in the 21st century - forget the word CBR :)

Between 90 and 100... probably not. Point was that I've had 75% encoded end up larger than the source. CBR is a quick way to get a predictable outcome. Sure, multipass VBR is the best but the 12+ hours vs 4 hours is a very real concern for many.