XviD presets thread

[aabxx]

(Note that I switched off AQ because at constant quantizer it really doesn't make any sense.)

Why doesn't it make sense?

It does make sense if you want base quant but also want some portions of the image which is more compressable 1 or 2 quant up from the base quant (much the same as with using b-frames which are 1 or 2 quants higher than the constant quant used).

[Teegedeck]

The question is: do you want to save 0,5% bitrate in exchange for the risk of blocking in dark areas of the picture (which I would call a degradation of quality)?

IMHO AQ only makes sense in two-pass where the 0,5% of bitrate are redistributed across the movie. And this only if we are very tight on available filespace. Also, I still have to figure out whether the risk of blocking is so big that I'll have to dump my beloved AQ for good, after I've put up such a long struggle... ;)

Seriously, if you encode at constant quantizer all that AQ gets you is some macroblocks with degraded quality and one or two MBs saved that you don't really have use for because constant-quantizer encodings have quite an unpredictable filesize anyway.

[G_M_C]

I've recently been using this matrix, and it gives fine sharp results on my stand-alone (Denon DVD-1920):

http://tweakers.net/ext/f/7c13fc31ddc8b5cd3de7c290e9a6e317/full.png

What's your opinion about this one ?

[Sharro]

I guess this is not a CQM discussion thread and doing it here would make presets enter on a loop.

Just my 5 cents.

All the best,

Sharro

[HeadBangeR77]

Isn't it Soulhunter's V3?
If so, then it's one of my favourites, but for certain uses, for instance it keeps filmgrain very well.

Could be a good alternative for the following presets: XviD>45% HQ or XviD>58% fast.

PS. But basically I must agree with Sharro - if you proposed the one for presets, with some more descirption etc. that would be fine.

[aabxx]

Teegedeck:
Actually I regularly save 10 to 20% bitrate with AQ turned on. I'm not doing any reencoding of mpeg-2 files like most here though, but mostly 80s VHS caps in xvid's interlaced mode.

I appreciate those 10-20% saved bits much although they of course come at the expense of some quality degradation. But with 1-pass encodings the only other way to save the same amount of bits is to use higher base quant/different matrix (or even higher b-frame quants but that would be just silly), and compared to those alternatives I'd much rather use AQ, as the quality degradation is supposedly happening in areas where quality matters less. At least I don't tend to experience blocking in dark areas and I concluded from my testing that AQ is doing a good job. Certainly, sometimes there's a little difference you can spot but it's far less than the difference would be if you had used one higher quant to begin with on all i/p-frames.

From limited experience I would also assume problems with blocking in dark areas would be much more common when transcoding from a lossy format, rather than encoding a lossless capture of an analog stream.

[Sharro]

Isn't it Soulhunter's V3?
...

Nope, Intra-Matrix is different. Should compress slightly better.

All the best,

Sharro

@HeadBangeR77, see my PM :)

[G_M_C]

Isn't it Soulhunter's V3?
If so, then it's one of my favourites, but for certain uses, for instance it keeps filmgrain very well.

Could be a good alternative for the following presets: XviD>45% HQ or XviD>58% fast.

PS. But basically I must agree with Sharro - if you proposed the one for presets, with some more descirption etc. that would be fine.

Actually ... i've taken this one from DVD-RB's Matrix editor addon. When you install the addd-on with DVD-RB, you can choose this one for use with CCE. It's the "High-Low" matrix (the "3rd best").

There are 2 matrices that are supposed to be "better" ("High-Medium" and "High-High"). One of them is equal to the one I posted, but has flat 8's in the inter-matrix (in stead of 16's).

I'm going to try that one in the next few days, but i predict that it won't work with my stand allone because of the 8's in the inter-matrix ... but you never know :P

I wouldn't even have guessed that the MPEG2 matrices would be acceptible with XviD ;)

[HeadBangeR77]

I wouldn't even have guessed that the MPEG2 matrices would be acceptible with XviD ;)
Now you know, and I know where you've taken the matrice from. ;)
I'm using Fox Home Entertainment occasionally, but it's really high bitrate for XviD encodes.

[Vitecs]

Hi all,

My SP does not like b-frames for encodes with high bitrates (>1500) and couple of audio tracks.

Will "XVID Presets" still have sense with disabled b-frames? (I do not mean "special purpose" presets.)

Thanks.

[Sharro]

Hi all,

I would like to have some feedback from the creators of the presets if this discussion of matrices and standalones should be taken into consideration here or not as I don't like to post off-topic matters and it could end in an endless loop for the presets.

@G_M_C: You better test with a small clip, I have a Philips DVP720SA (Mediatek1389EE) and a Philips5960/12 (still haven't found the Mediatek Chip Version).

The 5960 play's didees6of9 well but the DVP720 doesn't, also Yamada 6800 (1389DE) plays both well. I believe it has to do more with the ARM used in the firmware than a capability of the chip itself.

I don't want to test any more matrices otherwise my brain will be all messed up, HeadBanger can you find a way to send me a FOX matrix sample that I can test on both my players ?

All the best,

Sharro

[Teegedeck]

Hi Sharro,

I updated post 1 of this thread to explain why I don't think this thread is the right place for that, and also answered your PM to explain it in a bit of a lengthier fashion.

To summ it up: I'd prefer to keep it clearly separated; this thread is about purely quality-oriented presets, standalone-compatible presets (which are a legitimate and good idea) could be discussed in a separate thread or in the avs2qxvid.bat-thread because this batchfile already does something very much like what you asked for.

[professor_desty_nova]

Hi Teegedeck,

In the first post you say "1st pass @ quant=xy means that you have to set that quantizer as a zone in the first pass", but I only see it in the Anime preset (in the older thread they appeared in all presets). Am I blind, or the first pass quantizer zone settings aren't needed anymore?

[HeadBangeR77]

"First pass settings:
Same CQM as in second pass; same number of max consecutive b-frames, b-frame ratio and b-frame offset as in second pass; AQ if used in second pass; Trellis; ME precision=5, (except for the 'DVD-R perfect' preset, where it should be '4'); VHQ=1; Turbo. A zone with constant quantizer as specified in the preset-settings.
And that's it; no QPel, no GMC, no VHQ for-b-frames, no chroma ME."

Just above the main 2nd pass settings - I guess you were writing about this part.
So the same quantizer as used in the pre-set of your choice as the minimum value (usually Q3). Without a zone we would run a fast first pass, while the goal here is to run a mixed/hybrid 1st pass, sth between the fast one and the full quality one. It should work faster than the full quality option and ensure proper gathering of statistics, much more accurate than a fast first pass.

cheers,
HDBR77

PS. I've just reread the main post, and the way it's written it might sound confusing.

[ronnylov]

I wonder about the compressibility check. I was trying to use Enc to do a compressibility test but the program minimized and disappeared when starting the test (I had to stop the process in the task manager because there still was some disc activity going on). Maybe this is how it is supposed to be? Or maybe it did not like my avisynth script.

So I want to do a manually made compressibility check. What is the idea? Make a shorter sample of the original (like selectrangeevery in avisynth) and make a "full quality constant quant encode". The percentage value is that the target size compared to the size of the "full quality encode"?
Of course the target size is decreased to the same level as the original length was reduced (I mean if I encode 5 % of original video length to check compressibility, then full file size at "full quality" would be approximately 20 times higher).

Then I guess that by using 6 of 9 matrice and q=3 you get "full quality". Then choose the preset according to how much smaller in file size I want to make the final encode?

An example:
A sample of 5 % of the full video length gives a "full quality size" of 100 MB. My target file size is 1000 MB, and 5 % of 1000 MB is 50 MB. Then the compressibility check give 50 % compressibility because I need to compress my target to 50% of the full quality file size?

I have seen explanations like "target size compared to original size" but before encoding there is no real original size, just a bunch of uncompressed video frames that may have been originated from a original file with another level of compression...

[henryho_hk]

I suggest you trying Virtualdub with Enc (as I could not get it work with avs2avi). Actually, it is quite difficult to do manual "sampled" compressibility test because you need to sum up the individual frame sizes, discarding the extra I-frames (scene changes introduced by SelectRangeEvery) and its related b/p-frames. If Enc refuses to work, you can also try my batch script for Teegedeck's comptest.

Your example is correct. Compressibility test result is calculated with respect to source material, selected encoding parameters and a target size. In another sense, it is how "realistic" your target size is. Teegedeck's comp-test parameters serve as a well-defined, objective and consistent evaluation of the source material with respect to the target size. From the results, we can choose the appropriate encoding presets with predictable results (sharpness, details, etc). IMO, this is far better than the common "encode" - "blocky/undersized" - "complain" approaches.

[ronnylov]

Thank you henryho_hk for the explanation. I tried latest VirtualDubMod with Enc and also VirtualDub (not sure if it was latest version) with Enc but I had the same problem. It was an avisynth script with HDTV in 720P I tried.

EDIT: Moved question regarding the batch script to it's own forum thread.

[naylor83]

Does anyone know how to get the compression test settings to stick in Gordian Knot?

It will remember the settings for First Pass and Second Pass, but for each video I want to compress I have to edit the Xvid settings for the compressibility test. Slightly annoying.

[fight2win]

where to get latest build of xvid encoder to be used with vdub/enc?

[bur]

You're registered for one and a half years and still don't know where to get Xvid builds? :)

And while this isn't the proper thread for these questions, try koepi.org (http://www.google.com/search?q=xvid%20builds&hl=en).

[fight2win]

why aren't jawor's matrices used in megui presets?

[anubhavrocker]

This is a gr8 thread

[cyfaws]

I'm getting very weird results from the comp check, like 140%. Any ideea what would cause such abnormal values?

[Brother John]

That means your video is very “simple” to encode. You should reduce the target size, add audio tracks or anything else that eats space.

If you encode as is you’ll end up with a file smaller than your target size. But that doesn’t mean it’s bad quality.

[henryho_hk]

Comp. test result is target (sample) size divided by the compressed (sample) size. Hence, if you can accept a 70% "quality level", you can calculate your new target size as "old target size / 1.4 * 0.7".

[Buggle]

In the past I have done lots and lots of encoding, used to know which options to use which not, etcetera. After Xvid became final I got less interested in knowing everything and knowledge quickly decayed. I must say, after reading this thread my interests have sparked again!

I encoded a dvd episode using MPEG Matrix, but was not at all pleased with the result (although the quality should be very very high). I encoded, just for reference, a x264 crf19, finding out that is was about 180MB smaller. And still it looked better (sharper, less color bumping, etcetera). I played around a little with the settings, only to find no suitable answer to my needs. Encoding to 80%+ quality gives much other demands than queezing an encode on a CD...

Now that has led me to the interesting discovery that QPel actually made the file bigger. And as far as I know, QPel is supposed to bring the file size down if it was succesfull, so to speak.

And that, in turn, brings me to another question; how big is the room to play with in the presets? How much and which settings can be changed without breaking the presets value/goal?

As using Heini MR gives me a 141.3% compressibility with same settings and avs, whereas 6of9 gives only 81.3%. I would want to use 6of9 to retain the most 'truthful' picture, but cannot use the exact settings of the >90 preset.

Edit1: just finished the encode using the >90 preset even though I should not have used it. Picture looks absolutely horrible... I have the feeling 6of9 tries to encode too much of the noise present in the picture. There's so much mosquito noise around the edges, picture is totally 'flat'. Proof of the preset not being useful for <90 :P
Edit2: feel I am not really clear on what I mean, since >58 still uses 6of9. As I understand it correctly, 6of9 is able to retain the most detail in the picture, whereas other quant matrixes would give different results. What I mean to ask is this; could I use the settings in the highest preset together with, let's say, 6of9HVS? It feels a bit strange to go for >58 when in fact 80 is closer to 90 than 58. So for this reason I could still use 6of9 but change some little settings, keeping as much as possible of the 90 preset. But the question is, do I break anything doing that... and if so, which ones could easily be changed. My idea for instance would be to keep the CC settings (=off) but change to the B frame settings of >58.
Edit3: I've come to the conclusion that I do not like the results 6of9 gives me, even with 58 settings for an 80 clip. Going Q2 now with Heini, let's see what it will give me.

[henryho_hk]

Buggle, your findings are interesting. Would you please post some sample pictures here?

[squid_80]

The aim of QPel is to increase quality, not lower filesize.

[Buggle]

Buggle, your findings are interesting. Would you please post some sample pictures here?

Well, I think I have figured it out now, reading more about the Matrixes. I was comparing a normal Quant matrix (Heini MR) comptest @Q2 to a 6of9 also @Q2, which in fact should have been Q3. The comp tests are comparable now this way.

I am currently encoding it with V3HR, I like the results much better. I took a look at the source, only to see that it had some encoding errors in it already... Looks like someone messed up the encoding in the studio's. There were four points in the beginning of the ep (haven't seen more of it) that had some 'suspicious' pieces. One was for instance a face which had shimmering light on one side and shadow on the other. Looking at the encodes, the shadowside was very grainy/mosquito-ish, or differently said, plain ugly. That happend in other places as well. A different problem I noticed was little dancing colorplanes on the nose of Meredith Grey (yes, those episodes :p).

Checking back with the source, I found out that it was that to blame; but encoding it made it look worse, but even more with 6of9 than with Heini MR or V3HR. To get a result I am happy with would take me some filtering, but I do not have the time to get into that. For now, I'll settle with V3HR and sharp bicubic (compcheck around 77%). The source is just too 'unruhig', I normally always use lanczos.

[#2]

Hi there.

Possibly a silly question. I can't find where to get this profile. Can some one tell me please.

Thanks :)

[jeffy]

Hi there.

Possibly a silly question. I can't find where to get this profile. Can some one tell me please.

Thanks :)

I am not quite sure.
http://forum.doom9.org/showthread.php?t=107897
partially quoting the first post on the first page of the thread:
for editing, tooltip: 'fast, near lossless encoding, huge filesizes, NOT for archiving purposes': ME6, VHQ=1, no b-frames, chroma ME, EQM EHR , only for quality-based 1-pass, quantizer-restrictions constant quantizer 2

Can someone else please comment if this is right? :thanks:

[#2]

Yeah. It said you could download it from here. But I couldn't find it any where?

Any one got a copy or a link?

[henryho_hk]

The "editing" preset is in the 1st post of this thread.

[xbox360]

I am usig the cmd with xvid_encraw.exe, can you please help convert the folowing preset to work with CMD.

XviD '>90% comp. check' (HQ): MSP=4, VHQ=3, VHQ for b-frames, b-frames 1/1.00/1.00, QPel, b-frame sensitivity=-3, SixOfNine, no curve-compression, quantizer-restrictions min. 2, max. 4 (3 for I-frames), Overflow Control Strength=0

[Buggle]

I am usig the cmd with xvid_encraw.exe, can you please help convert the folowing preset to work with CMD.

XviD '>90% comp. check' (HQ): MSP=4, VHQ=3, VHQ for b-frames, b-frames 1/1.00/1.00, QPel, b-frame sensitivity=-3, SixOfNine, no curve-compression, quantizer-restrictions min. 2, max. 4 (3 for I-frames), Overflow Control Strength=0

You might want to consider using MeGUI (www.x264.nl) which has the presets build in.

[henryho_hk]

Teegedeck, should we withhold using the Chroma Optimizer option until the bug found by Plugh at http://forum.doom9.org/showthread.php?t=136324 is fixed?

[Sharktooth]

i added some new presets to megui.
two HQ 2 pass presets (one with Qpel and one without it) and 2 HQ quant 2 presets (one with Qpel and one without it).
hope you'll like them.

[elguaxo]

i added some new profiles to megui.
two HQ 2 pass profiles (one with Qpel and one without it) and 2 HQ quant 2 profiles (one with Qpel and one without it).
hope you'll like them.

I just updated all the profiles but the CQMs point to a directory that doesn't exist:
C:\Documents and Settings\Ebola\Local Settings\Temp\nd5bvkgo.myj\extra\Didees SixOfNine.cqm

[Sharktooth]

the CQMs are not included. you have to provide them manually.

[elguaxo]

the CQMs are not included. you have to provide them manually.

I know, but the path is confusing. I know I never had a C:\Documents and Settings\Ebola\ dir :p

[Sharktooth]

cant do much right now.
we'll find a way to include the CQMs or we'll remove all profiles using them.

[rack04]

i added some new profiles to megui.
two HQ 2 pass profiles (one with Qpel and one without it) and 2 HQ quant 2 profiles (one with Qpel and one without it).
hope you'll like them.

Am I correct that these new profiles don't require a compressibility test? Anyone care to explain these two new profiles? Thanks.

[Sharktooth]

yes you're correct.. they do not require it and i added them on purpouse, coz megui still dont have an "internal" comp-test.
as i said they're 2 2-pass bitrate based profiles and 2 constant quant 2 profiles for maximum quality regardless of the bitrate/filesize.

[rack04]

With the comp check XVID profiles I'm now getting the following error:

[NoImage] Standard error stream: xvid [error]: Error opening matrix file .\extra\Didees SixOfNine.cqm

[Sharktooth]

you have to point megui to the matrices location for every comp.check preset and re-resave them (update button) as for a new installation.

[rack04]

you have to point megui to the matrices location for every comp.check profiles and re-resave them (update button) as for a new installation.

Did something change in the profiles because I thought this was fixed already? Maybe not. Thanks for the help.

[Sharktooth]

no, i restored the deafault path which is .\extra\etc..... which is however incorrect coz .\ is not valid.
do what i said and re-add the CQMs to the presets.

[rack04]

do what i said and re-add the CQMs to the profiles.

Yes sir. :D

[Sharktooth]

sorry for the imperative but there's no different solution unless you restore the old presets.

[rack04]

sorry for the imperative but there's no different solution unless you restore the old profiles (bad idea).

I'm fine with your solution. Just giving you a hard time. :p