XviD presets - preliminary thoughts and suggestions

[foxyshadis]

You can't. "xvid" is a VFW tag, not a mkv tag, and encraw cannot write vfw mkvs. (Doesn't xvid have a "generic mpeg4" decoding option though? ffdshow does.) If you really must use it, you have to encode to avi and mux that to mkv (without --engage native_mpeg4).

[loro]

Teegedeck you're right about the preset selection. But the quantizer maximum value was 5. I will try again, if not I'll use >30% HQ preset. I'll let you know.

Also I loaded my stats file into StatsReader and it shows 2122 MB. Result size is 1726 MB.

[Teegedeck]

First pass at quantizer=3?

[loro]

Yes. Fixed quantizer=3. My compressibility check in Enc was 45% with SixOfNine qantization matrix. So I also tried compressibility check with sharktooth's EQM v3HR quantization matrix and the result was 57%.

[Teegedeck]

This is a puzzler. If 1st-pass-filesize at quantizer=3 was 2122 MB, then it couldn't have been the quantizer restriction that prevented the filesize from going below 1726 MB. (At least it's pretty unlikely; for example I have a movie here where this preset yielded a first pass filesize of 2437 MB and at quantizer 4 it would have been 1734 MB. I would think that quantizer 5 should have brought it down at least solidly under the 1500-MB-margin. But of course my base of comp. check data is not that big, yet).

Can you perhaps make a filesize prediction with Enc using the >45% (HQ) preset at constant quantizer=5 in order to verify this? And maybe another one using the >30% (HQ) preset?

Maybe curve-compression plays a role here - how quant-restrictions and curve-compression interact is still largely untested, I believe.

[shpitz]

i've done a test encoding of a 1min clip from a 720p 60fps capture.

i've tried all the profiles henry_ho set in his batch file.

i've got a dual xeon 3ghz.

all values are fps (rounded).

http://img56.imageshack.us/img56/2697/encodingspeedcomparison0sn.th.gif (http://img56.imageshack.us/my.php?image=encodingspeedcomparison0sn.gif)

all profiles produced the same exact filesize (2 threads vs 4 threads) except dvd_perfect and slow_medium.

dvdr_sharpen was unplayable, it crashes every player i tried.

[henryho_hk]

The following is the BAT script implementing the new presets (cmprgt30fast, cmprgt30hq, cmprgt45fast, cmprgt45hq, cmprgt58fast, cmprgt58hq, cmprgt90fast & cmprgt90hq). This version does not have batch mode support and you can only feed it with one AVS at one time.

If you specify "cmprchkfast" or "cmprchkhq" as parameter, it will fire up Enc for you to perform a comp test (refer to the past posts for details). Since I have not found any way to automate Enc yet, you need to click a few buttons and close Enc manually. The BAT will presume your comp test is successfull and parse the log file for the result so as to pick the right preset (fast/hq as you specified) for your AVS.

Before you use the BAT script, remember to set the full path of Enc and Virtualdubmod at the start of the script.

Please enjoy.

-----------------------

20060712: Corrected some stupid mistakes in comp test result parsing.
20060808: Unofficial ESS and MTK presets added.
20060809: Preset ">45% Fast" corrected
20060814: BAT script clean-up
20061004: obsolete old code deleted
-----------------------

[Teegedeck]

Smokin'! :cool:

I hope to get some computing time on our Windows box tomorrow so I can try it out.

[shpitz]

i've installed win2k3 x64 enterprise and will play with it and see how it works on x64...

i'm using celtic druid's latest x64 build of xvid.

not sure about avisynth, does anyone have a link to a x64 build of it?

[henryho_hk]

http://extended64.com/blogs/wiak/archive/2005/09/28/1368.aspx

Please visit the AVISynth forums for more details.

[shpitz]

http://extended64.com/blogs/wiak/archive/2005/09/28/1368.aspx


thanks, already been there, all links are dead ;-(

i have d/l'ed a while ago the builds, i'm playing with them right now...

i'm using your scripts henry, will report back with some data.

EDIT: enc doesn't seem to work with x64, none of the codecs to the right is available (all grayed out) and when i click configure i get an error message "ICOpen failed"

just in case some1 else needs x64 builds: x64 builds (http://okejl.dk/dunstan/)

[henryho_hk]

My to-do list:

1) Handle <30% comp test cases. Will there be a "10%" preset, Teegedeck? :devil:
2) Get rid of the goto hacks with in for-loop.
3) Implement audio muxing.
4) Implement batch processing.
5) Implement some unofficial 1-pass presets. [I suffer from a lack of patience. :devil: ]

As for Enc under x64, I don't think it will work, as it depends on avs2avi, which in turn depends of VFW, which probably isn't available under x64.

xvid_encraw seems to access xvid.dll directly. Pls correct me if I am wrong.

[foxyshadis]

At <30% you might as well flag an error and state "resize and/or filter to bring above 30%!" If you don't want to be that drastic at least make it a warning. You'll definitely have to raise the max quant and kiss details goodbye either way.

[shpitz]

xvid_encraw works great in x64 and actually i got a reduction in speed when using latest xvid build (non x64) and the latest x64 compile by squid_80.

[Teegedeck]

My to-do list:

1) Handle <30% comp test cases. Will there be a "10%" preset, Teegedeck? :devil:
When there's XviD AVC and if it's still one binary.

You'll definitely have to raise the max quant and kiss details goodbye either way.No way, these are quality presets, not baddity presets! ;)

[Deano123]

Thanks For This Really Helps

[henryho_hk]

For quality freaks here. If you wanna deinterlace, remember to use tdeint(). It's slow, significantly slower than leakkerneldeint(). But then its quality is much better, being sharp while leaving far less residual comb. It usually allows me to use at least one higher level of preset than leakkerneldeint().

[Teegedeck]

Could someone please test something for me before I change my attachment to 'thre relevant post about the presets? : When I load the zip with those XviD profiles into MeGUI there seems to be nothing to import.

I've only changed 'EncodingMode' to '4' (aut. 2-pass). Maybe a problem with the archive?

[lilhobo]

there should be a wiki or knowledgebase damned hard to read thru all the posts lol

[Valeron]

Most anime-centred people here swear by H.263. Of course in the higher-bitrate areas one could use sharper matrices. Didee once was so nice to send me a draft of one. But you need a higher bitrate with such matrices in order to avoid ringing. My impression was that anime fans don't go for high-bitrate encodes. If there's demand and someone to test all that, it could be done. But I don't feel like it, I'm still not finished comparing fast modes. Perhaps we should have a dedicated film-grain-preserving mode, that's what I'm concerned with ATM.

hi, Teege, could you plz mail me that matrices?
i'm interesting in testing it myself.

[kurt]

Could someone please test something for me before I change my attachment to 'thre relevant post about the presets? : When I load the zip with those XviD profiles into MeGUI there seems to be nothing to import.

I've only changed 'EncodingMode' to '4' (aut. 2-pass). Maybe a problem with the archive?
importing the xvid profiles works fine here :)

but where I have to save the "extra folder" (with the cqm)? Megui produces an error when saving them into MeGui root folder... (same with profile folder)#

Edit: sorry I misunderstood. Importing the profiles via File Menu doesn't work :o

[Teegedeck]

Megui produces an error when saving them into MeGui root folder... (same with profile folder)#

Edit: sorry I misunderstood. Importing the profiles via File Menu doesn't work :oI was afraid of that, thanks for confirming it. So, when I have time again I'm gonna try to find the guilty part.

[elguaxo]

Inside the .ZIP there is a folder /XviD-profiles and inside it the folders /extra and /profiles. In the .ZIP with your previous profiles there was no /XviD-profiles folder.

[Teegedeck]

Gosh, simple as that? :o Thanks, gonna re-upload that zip tomorrow.

[Teegedeck]

hi, Teege, could you plz mail me that matrices?
i'm interesting in testing it myself.
Didée says it's OK, so you can have a look.

BIG, fat DISCLAIMER:
Didée put that matrix together spontaneously and didn't intend to publish it; he didn't test it at all. Thus it comes with a 'do-with-it-what-you-want-but-don't-blame-me'-license, and it has no name. I've used it a few times and found it produces nice results but only at low quantizers, of course. It's not meant for all kinds of anime, but only for very sharp and very clean sources. It could be used as a starting-ground for further anime-CQM development. If you want to do that you'd certainly be welcome.

16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 17
16 16 16 16 16 16 17 19
16 16 16 16 16 17 19 22
16 16 16 16 17 19 24 32
16 16 16 17 19 24 36 40
16 16 17 19 22 32 40 48

[Teegedeck]

OK, updated the XviD-profiles for MeGUI here (http://forum.doom9.org/showthread.php?p=841263#post841263).

[shpitz]

thanks Tee, will try them tonight.

[Teegedeck]

...nothing changed there except for correcting the encoding mode from 'CQ' to 'automated 2-pass'. ;-)

[firered]

Hey Teegedeck

Based on your SixofNine matrix for high bitrates I was wondering do you have a highlow or a medium to low bitrate matrixs for xvid or anyone can recommend a custom one?

thanks

[Teegedeck]

Hi, Firered,

SixOfNine is Didée's, not mine. And please read the main posts of the thread (they're linked at the beginning).

[sarmin]

how do i add the profiles?
and how do i add a matrices?

[Livesms]

how do i add the profiles?
and how do i add a matrices?


And how can I add custom matrix (SixOfNine for ex.) in MeGui?
Or it can be done only from a commandline.

[Teegedeck]

Simply download the MeGUI profiles I provided and use MeGUI's profile import function.

This will import the matrices as well. Afterwards you can load all imported matrices from the zones tab of MeGUI's config.

[JoeBG]

Simply download the MeGUI profiles I provided and use MeGUI's profile import function.


Sorry for the maybe stupid question: Where can I get the matrices?

[Teegedeck]

@sarmin: Don't pollute threads with basic questions. You already posted a thread for that topic

@kurt: Actually it's not only about MeGUI profiles but about XviD presets.

@JoeBG: There's three ways to get them: XviD VfW installs them into the XviD directory; they are part of the MeGUI profiles that I attached to the main post; LigH's CQM editor also comes with a vast collection of matrices.

[Livesms]

Simply download the MeGUI profiles I provided and use MeGUI's profile import function.

This will import the matrices as well. Afterwards you can load all imported matrices from the zones tab of MeGUI's config.

First of all thank's for your great job.
I've got several questions
1. I have troubles using Enc for compressibility test. I've discribed it in this thead http://forum.doom9.org/showthread.php?p=857491#post857491
2. Your MeGui profiles use only eqm_v3ulr_rev3, eqm_v3lr, eqm_v3hr, Didees SixOfNine matrix. And what about other ones? If i want to use other matrix how can i import it into MeGui.
3. Is there any reason to test/find/look for better quality then i can get using Enc test -> selecting profile according to %result with mark HQ.
4. I can't understand link with %result test and matrix you use in you profile. For example I've got 48%. So i can use XviD > 45% comp. check' (fast) or XviD > 45% comp. check' (HQ). But there is a different matrixe there eqm_v3lr.xcm and eqm_v3hr.xcm. And bitrate is also here. Or maybe bitrate change in Enc will cause % result change from one level to another. For example I test loing on 1200kbit/s and if I change it to 700kbit/sec will I get smth like 35%? I can't understand how can I use the same matrix for 720*576 DVD clear video and 320*240 port. camera video if it will return the same test results for the rather different bitrates (1400 and 600 for examp.)?

[Teegedeck]

Sorry, I'm not an expert on Enc. :(
CQMs belong into the 'extra' directory'; I believe it was at \profiles\video\extra. But you can load a matrix from anywhere you like.
If the result is just on the border between two profiles, it's well possible that both can be used and which profile looks better might depend. Otherwise: no, it doesn't make sense.
The Enc test is only there to prevent under- or oversizing in 2nd pass.
Clever question. :) I've found out during testing that the higher-quality EQMv3 HR matrix with HQ XviD settings (VHQ=4) produces about the same bitrate as the lower quality EQMv3 LR matrix with faster XviD settings. It's a good example of how using more efficient settings can visually translate into higher quality: The safed bits allow for using a matrix with lower coefficients. I hope to do something like that for all presets eventually.


Edit: I think I haven't fully understood question 4; can you please try to put it another way? Yes, Enc results depend on a lot of things; bitrate, video content, resolution. An HDTV still scene might yield the same percentage as a 320x music video at the same bitrate. Identical percentage would also mean that they would come out the same filesize if you encoded them with the same settings at the same constant quantizer. Is that what you meant?

@kurt: I moved your answer to sarmin's VDub question to sarmin's matrix-dedicated thread.

[Livesms]

1. Sorry, I'm not an expert on Enc. :(
I've solved this problem. It was all up to AviSynth - ordinary reinstall and all problems gone :) May be there was some bug in script
AviSource("Video.avi",audio=false) :D

2. CQMs belong into the 'extra' directory'; I believe it was at \profiles\video\extra. But you can load a matrix from anywhere you like.
Yes. But how can i tell Megui to load another matrix. I can only read command line generated by MeGui using entered settings/profile but not to change it directly. And I didn't find how can I manualy select matrix right from MeGui Xvid setting dialog.

3. If the result is just on the border between two profiles, it's well possible that both can be used and which profile looks better might depend. Otherwise: no, it doesn't make sense.
The Enc test is only there to prevent under- or oversizing in 2nd pass.
I was asking this question thinking of not only MeGui + Your Xvid Profiles. I'm interested is it possible to get better quality using other settings/profiles/matrix's (not only your's) then it will be when I simply run Enc test, select profile according to % result with HQ mark and enter bitrate I'm aiming.

4. Clever question. :) I've found out during testing that the higher-quality EQMv3 HR matrix with HQ XviD settings (VHQ=4) produces about the same bitrate as the lower quality EQMv3 LR matrix with faster XviD settings. It's a good example of how using more efficient settings can visually translate into higher quality: The safed bits allow for using a matrix with lower coefficients.
And how it can be found out? Think there is impossible to test video for each matrix for whole length with different Xvid settings to find out best balance quality/speed. It has no sense - too much time. But I'm not Nostradamus to see two step forward, what combination of matrix (maybe not HQ/HR with best quality) will give the same (best i can reach here) in less time/xvid switch on.

[Livesms]

Edit: I think I haven't fully understood question 4; can you please try to put it another way? Yes, Enc results depend on a lot of things; bitrate, video content, resolution. An HDTV still scene might yield the same percentage as a 320x music video at the same bitrate. Identical percentage would also mean that they would come out the same filesize if you encoded them with the same settings at the same constant quantizer. Is that what you meant?

Nearly :)

I can't fully understand means of this % percentage from Enc. What does they mean. May be fully understanding their meaning I will find answer to my quetion myself :)

[Teegedeck]

The settings I gave for Enc simply represent a filesize as big as XviD could get at maximum-super-duper-hyper-quality. :) These settings wouldn't produce maximum-super-duper-hyper-quality, of course, but the filesize would be about as big and the settings are fast.

Now Enc makes a filesize prediction about how big your video would come out encoded at that quality and then calculates to how much percent of that filesize the filesize you aim at amounts. If you look closely at the Enc results you'll discover that it always show you a ratio of the filesize you gave for the comp. check and another value - this value is actually the filesize prediction based on the XviD settings.) But we only need that percentage value as a means of comparison. With Enc we can compare the percentage that an encoding with a certain preset - telling from experience - typically produces to the percentage your two-pass-encoding to a certained filesize will probably produce. That way we can indirectly guess with which preset your aimed-at filesize could be reached.

If you test the presets at CQ with their maximum and minimum allowed quantizers (restrictions) you will see that an encoding at the maximum allowed quantizer of a preset will produce a filesize that represents a percentage of filesize Enc has calculated that is slightly lower than the percentage in the preset's name's. And at the minimum allowed quantizer of a preset it will produce a filesize that represents a percentage of the Enc filesize prediction that is quite a bit higher than the percentage given in the next-higher preset's name.

The Enc comp. check is there to prevent you from hitting the hard barrier of the presets' max./min. allowed quantizers.

Edit: I found this rather hard to explain, I hope it is comprehendable.

[Livesms]

Teegedeck

Ok. Then tell me why we run Enc comp test using settings you recommended CQ=3(or 4), Bvops 2-1.62-0, Qpel(-), GMC(-), AQ(-), Trellis(+), ME(4), VHQ(0-off), VHQ for Bf(-), ChromaME(-), Turbo(+).

Of course I will finally encode with ME(6), VHQ(1 or 4), ChromaME and other things.
As I understand Enc gives percentages results calculated as
%Percent = (DesiredSize) / (EncodedSize) * 100%

where
DesiredSize = Length of Input File * Bitrate / 8;
EncodedSize = FileSize encoded with Enc settings (Q=4, Bvops 2-1.62-0, Qpel(-), GMC(-), AQ(-), Trellis(+), ME(4), VHQ(0-off), VHQ for Bf(-), ChromaME(-), Turbo(+)).

So why should we compare file encoded with one setting to another? Look's like comparing what it more dark - green or red :) Or smth like that?

May be I'm wrong and my questions look's stupid/silly as a two-bob watch, but i can't realize meaning of this Enc %.

In jonny's site was description, reading which I understand, that Enc % is value represents visual quality, I will get compressing my video with entered bitrate (100% means visual identical video, 0% means absolutely different video). May by this scale is not linear, and may be logarithmic (it will be better - 50% means not a half of qual)

I've run test with first file fall into my hands and get smth like 126%. And that's make confused.

[Teegedeck]

Your result of over 100% means that you are using a way-too-small resolution for the filesize you aim at.

The Enc test relies on the assumption that encodings at fast settings somehow relate to encodings at HQ settings. XviD's fast first pass also relies on that assumption. Or how do you think a second pass can be calculated based upon a first pass that doesn't use GMC, doesn't use QPel, doesn't use VHQ=4 (but does use Turbo) - although you have selected all these settings for the second pass? It works because the results - although very different - somehow relate.

And testing seems to strengthen that assumption that the results of the presets-Enc test relate well enough to the real-life results of preset-2-pass-encoding.

Let me put it like this (excuse me if I simplify the actual values for the sake of making it easy): A movie encoded with the '>45% HQ' preset at constant quantizer 4 will come out half as big as the same movie encoded with the very fast settings I gave for the Enc test. A movie encoded with the '>30%' preset at constant quant=4 will come out 30% the filesize of the same movie encoded with the Enc preset. And so on. And all we need to learn from performing the Enc test is whether or not the average quantizer of our preset-encoding would come dangerously near the maximum/minimumallowed quantizer for that preset.

What matters is that the results of the settings relate, not that they are similar. I could have chosen a completely different constant quantizer for the Enc test, so that (for example) the '>45%' preset would have been a 120% preset. But I thought that this would have been confusing. A higher-quality setting would have had similar (though less dramatic) consequences, plus the disadvantage that the test would take much, much longer. So I chose a quantizer that would make it clear that something's not okay if you get more than 100% and fast settings.

It isn't scientific, it's just a rule of thumb, made to spare you making a blind guess. I'd say, 9 out of 10 times it should work.

[Livesms]

Your result of over 100% means that you are using a way-too-small resolution for the filesize you aim at.

The Enc test relies on the assumption that encodings at fast settings somehow relate to encodings at HQ settings. XviD's fast first pass also relies on that assumption. Or how do you think a second pass can be calculated based upon a first pass that doesn't use GMC, doesn't use QPel, doesn't use VHQ=4 (but does use Turbo) - although you have selected all these settings for the second pass? It works because the results - although very different - somehow relate.

And testing seems to strengthen that assumption that the results of the presets-Enc test relate well enough to the real-life results of preset-2-pass-encoding.

Let me put it like this (excuse me if I simplify the actual values for the sake of making it easy): A movie encoded with the '>45% HQ' preset at constant quantizer 4 will come out half as big as the same movie encoded with the very fast settings I gave for the Enc test. A movie encoded with the '>30%' preset at constant quant=4 will come out 30% the filesize of the same movie encoded with the Enc preset. And so on. And all we need to learn from performing the Enc test is whether or not the average quantizer of our preset-encoding would come dangerously near the maximum/minimumallowed quantizer for that preset.

What matters is that the results of the settings relate, not that they are similar. I could have chosen a completely different constant quantizer for the Enc test, so that (for example) the '>45%' preset would have been a 120% preset. But I thought that this would have been confusing. A higher-quality setting would have had similar (though less dramatic) consequences, plus the disadvantage that the test would take much, much longer. So I chose a quantizer that would make it clear that something's not okay if you get more than 100% and fast settings.

It isn't scientific, it's just a rule of thumb, made to spare you making a blind guess. I'd say, 9 out of 10 times it should work.

As I can see your profiles was named according to Enc result with Q=4 (and other recommended). So if I get 120% I can lower bitrate or allow bigger frame size. Ideally is to get 100% (>90%) or I'm wrong?

But

With the compressibility test function you are able to predict the quality of your encode, before making the real encode.
The result is a % value, to obtain good results you should stay in the 50%-60% range (the range is not objective, you should make experiments to see what you like more).
If the result is too low, you can change resolution, resize filters, codec parameters, use more cds in order to obtain an higher value.
--------------------------------
http://jonny.leffe.dnsalias.com/enc/htm/htm/xvidct.htm

According to this - the best way is to go for 50-60%.

And another quest. You use different matrix for fast and HQ profile with the same %
for XviD '>45% comp. check' (fast) - eqm_v3lr
for XviD '>45% comp. check' (HQ) - eqm_v3hr

But EQM V3HR - This is a HIGH BITRATE matrix designed to work at full anamorphic DVD res (720x432) with bitrates over 1600kbits (for an average motion movie) and works pretty well even with Q2 encodings or extremely High bitrates. But for that target there will be an Ultra-high bitrate matrix (EMQ V3UHR).
And EQM V3LR - This is a LOW BITRATE matrix designed to work at full anamorphic DVD res (720x432) with bitrates lower than 1500kbits (for an average motion movie) and works pretty well even with high motion sources (using higher bitrates) keeping details where other matrices produce artifacts.

Looking at that I can't understand how it can be. You said proc% do not directly depends on bitatre. I can get 50% with 720*576 DVD-9(8Gb) video for bitrate 1800kbit/s and the same perc% for 320*240 digi cam video for bitrate 600kbit/s. And i will use eqm_v3lr(LOW BITRATE matrix) for fast profile in both case - for DVD with 1800kbit/s and Digi cam video with 600kbit/s. And the same with HQ profile - I will go with eqm_v3hr (HIGH BITRATE) either for DVD with 1800kbit/s and DigiCam with 600kbit/s.

How can explain me that :)

[Teegedeck]

As I can see your profiles was named according to Enc result with Q=4 (and other recommended). It was Q=3; that might also explain your high percentage result.

So if I get 120% I can lower bitrate or allow bigger frame size. Ideally is to get 100% (>90%) or I'm wrong?Ideal is a full resolution. The percentage is secondary if you want to weigh it out.

But


http://jonny.leffe.dnsalias.com/enc/htm/htm/xvidct.htm[/i]

According to this - the best way is to go for 50-60%.
Please forget that; it has nothing to do with the XviD presets.

And another quest. You use different matrix for fast and HQ profile with the same %
for XviD '>45% comp. check' (fast) - eqm_v3lr
for XviD '>45% comp. check' (HQ) - eqm_v3hr

But EQM V3HR - This is a HIGH BITRATE matrix designed to work at full anamorphic DVD res (720x432) with bitrates over 1600kbits (for an average motion movie) and works pretty well even with Q2 encodings or extremely High bitrates. But for that target there will be an Ultra-high bitrate matrix (EMQ V3UHR).
And EQM V3LR - This is a LOW BITRATE matrix designed to work at full anamorphic DVD res (720x432) with bitrates lower than 1500kbits (for an average motion movie) and works pretty well even with high motion sources (using higher bitrates) keeping details where other matrices produce artifacts.Haven't I explained that above? I believe I have.

Looking at that I can't understand how it can be. You said proc% do not directly depends on bitatre. I can get 50% with 720*576 DVD-9(8Gb) video for bitrate 1800kbit/s and the same perc% for 320*240 digi cam video for bitrate 600kbit/s. And i will use eqm_v3lr(LOW BITRATE matrix) for fast profile in both case - for DVD with 1800kbit/s and Digi cam video with 600kbit/s. And the same with HQ profile - I will go with eqm_v3hr (HIGH BITRATE) either for DVD with 1800kbit/s and DigiCam with 600kbit/s.

How can explain me that :)Bitrates are a sham. That is, if you only look at bitrate and not at resolution and content. Reduce resolution enough and you can encode anything with a high-bitrate matrix at 600 kb/s. Edit: There are DVDs that come out at 3.5 GB if you encode them with SixOfNine at quant=3 and there are DVDs that come out 1.5 GB at the same settings. Does it make sense to use the same CQM on both movies just because you aim for the same filesize/bitrate for both movies? No.

[Livesms]

Ideal is a full resolution. The percentage is secondary if you want to weigh it out.
Why? Is it better encode blocked full resolution video than resized but better encoded?

According to this - the best way is to go for 50-60%.
Please forget that; it has nothing to do with the XviD presets.
?????? Can you explaine?

Haven't I explained that above? I believe I have.
Can you give me a link or some more explanation?

Bitrates are a sham. That is, if you only look at bitrate and not at resolution and content. Reduce resolution enough and you can encode anything with a high-bitrate matrix at 600 kb/s. Edit: There are DVDs that come out at 3.5 GB if you encode them with SixOfNine at quant=3 and there are DVDs that come out 1.5 GB at the same settings. Does it make sense to use the same CQM on both movies just because you aim for the same filesize/bitrate for both movies? No.
Why should I encode with HighBitrate matrix with 600kbit? I just interested why you use the same(different) matrix proceeding only from perc% and from bitrate? In your profiles matrix depend on Perc but on bitrate.

[shpitz]

poor Teegedeck, what did you get urself into... lol

[Teegedeck]

I'm afraid I seem to fail miserably when it comes to explaining things. :o There comes a time when one just has to admit defeat.

[henryho_hk]

And i will use eqm_v3lr(LOW BITRATE matrix) for fast profile in both case - for DVD with 1800kbit/s and Digi cam video with 600kbit/s. And the same with HQ profile - I will go with eqm_v3hr (HIGH BITRATE) either for DVD with 1800kbit/s and DigiCam with 600kbit/s.

How can explain me that :)

May I share some of my experience in XviD encoding.

When Teegedeck said comp% does not directly depend on bitrate, he means that comp% is not in a simple linear relation with the bitrate. It also depends heavily on the matrices, other encoding options and quantizer limitations imposed on XviD encoder (in 2-pass mode). For example, in the >45% Presets, in order to compensate for the HR matrix, we need to use the VHQ mode 4 and thus much more encoding time to achieve the same bitrate.

IMO, the Enc comp test is actually an assessment of the source material helping us to choose a right preset so as to achieve an intended bitrate. After obtaining the Comp% value, you are still given one more choice: getting a good encode at a higher speed, or getting a better-to-excellent encode at a slower speed. That's the difference between the Fast & HQ presets. The other preset pairs look much more alike but you can still see the speed-quality compromise.

[Livesms]

May I share some of my experience in XviD encoding.

When Teegedeck said comp% does not directly depend on bitrate, he means that they are not in a simple linear relation, not to mention all other XviD parameters, matrices, etc. When he said bitrates are a sham, he means that the bitrate should not be not an input parameter; rather, it is an result of the resolution you chose, the quality you desired (filters, matrices, etc.) and the encoding time you are willing to spend. If you give XviD more time, it can make more efficient encodes, giving you sharper & smooth movie at the same bitrate. On the other hand, when you are out of time, please stay at lower quality presets or use the higher quality "fast" preset (and be prepared to obtain a big file).

The goal of Enc comp. test is actually to suggest us a preset based on the bitrate we specified. Then for the "fast" or "HQ", it is a judgement based on our available time. It does have an impact of quality, but hey, I did not recall Teegedeck had ever told us that ">58% Fast" has a better quality than ">45% HQ".

It's the beauty of XviD - flexibility and versatility.

:goodpost: :thanks: :goodpost:

Ok. I just started to realize meaning and all thins of Enc+Xvid shown in Enc perc%. I'll read one's more main therad postes.
Another question - is it possible to get better quality then I will get using Enc -> MeGui with selected profile?

[henryho_hk]

is it possible to get better quality then I will get using Enc -> MeGui with selected profile?

Oops, you caught my original version. ^_^

Actually, I have another question for Teegedeck. Does MeGUI supports chroma optimizer now in the latest version? If it does, my BAT script shall retire. ^_^