Auto Gordian Knot: current version 2.55

[richarddd]

I was just thinking about these issues, came to the forum, and there was your post :) I have a Philips dvp642, btw.

I've been coming to the conclusion that a lower quality encode looks better on my SAP and (non high-def) TV. For a typical movie, I can easily tell the difference between a 1 cd and 2 cd encode on my computer, but can't tell the difference on my TV, other than that a 2 or higher cd encode is more likely to stutter on the TV.

I wonder if (1) burning to DVD will work better than burning to CD (I don't yet have a DVD burner) and (2) quality differnces would be more apparent on a high definition TV than a regular TV.

What is not implemented in AGK is the high bitrate limiter option included in the 2nth pass of the Xvid 1.1 encoder. From what I understand it should be the answer to extreme situations.

Sounds like the best solution. Is there a way to fake it with AGK?

I suppose AGK won't support xvid 1.1 until there is an actual 1.1 :D

[BigDid]

Originally posted by richarddd
I was just thinking about these issues, came to the forum, and there was your post :) I have a Philips dvp642, btw.
It should be like the 630 with a ESS Vibrato chip; as you have'nt stated it -> best use the .ess option of AGK
... other than that a 2 or higher cd encode is more likely to stutter on the TV.
If it is related to the high birate, your SAP is more sensitive than mine -> try Xvid 1.1
I wonder if (1) burning to DVD will work better than burning to CD (I don't yet have a DVD burner) and (2) quality differnces would be more apparent on a high definition TV than a regular TV.
I was given the info that DVD is better/much better? than cd for video streams, I do use DVD. For the HDTV, I've read somewhere else on the forum one or more thread related to this, it seems the answer is yes related the the pixel width of the HDTV ?...
What is not implemented in AGK is the high bitrate limiter option included in the 2nth pass of the Xvid 1.1 encoder. From what I understand it should be the answer to extreme situations.
Give a try to Xvid 1.1 with the default parameters for a 15 to 20% improvments on the high bitrates.
Sounds like the best solution. Is there a way to fake it with AGK?
I suppose so, havent tried it myself with
-AGK tweaker -> needs to confirm the changed scripts for the avisynth part and vdmod 3 or 4 times each, you do it for 1 movie after you may grow impatient;)
-AGK expert -> -New- more or less like the first one have not used it, see the related thread
-Just read the post above mine from CeeJay.dk and I may give it a try; the difference with the above solutions (if I understand it well) is that you pause one time just at the beginning of the first pass, make your changes in the script ( for filtering) change the encoder options via Vdmod -to make those changes persistent through the second pass- and resume; up to AGK to re-run the first pass and adjust the parameters if needed... in theory ;)
I suppose AGK won't support xvid 1.1 until there is an actual 1.1 :D
No support does'nt mean it does'nt work;) In fact I have'nt gone back to 1.02 since my tests, but if problems ... etc...

Geee 2 long typings in english for one day is good but exhausting, hope that helps, read you later:) .

Did

[richarddd]

Originally posted by BigDid
It should be like the 630 with a ESS Vibrato chip; as you have'nt stated it -> best use the .ess option of AGK

Of course :)

If it is related to the high birate, your SAP is more sensitive than mine -> try Xvid 1.1

It often stutters on 2 CD encodes on scenes consisting mainly of moving leaves or rain or the like.

I was given the info that DVD is better/much better? than cd for video streams, I do use DVD.

I read in the hardware player forum that players read DVDs at a higher bit rate than CDs. On the other hand, why not just back up the DVD (or the main movie) uncompressed if you have a DVD drive?

For the HDTV, I've read somewhere else on the forum one or more thread related to this, it seems the answer is yes related the the pixel width of the HDTV ?...

I would have thought HDTV has more pixels than regular TV and therefore shows more movie detail.

Give a try to the default parameters of the Xvid 1.1 for a 15 to 20% improvments on the high bitrates.

What do you mean by a 15-20% improvement?

I suppose so, havent tried it myself with
-AGK tweaker -> needs to confirm the changed scripts for the avisynth part and vdmod 3 or 4 times each, you do it for 1 movie after you may grow impatient;)
-AGK expert -> -New- more or less like the first one have not used it, see the related thread
-Just read the post above mine from CeeJay.dk and I may give it a try; the difference with the above solutions (if I understand it well) is that you pause one time just at the beginning of the first pass, make your changes in the script ( for filtering) change the encoder options via Vdmod -to make those changes persistent through the second pass- and resume; up to AGK to re-run the first pass and adjust the parameters if needed... in theory ;)

Seems a fair amount of effort is required. Especially for those of us who use AGK because we don't really understand how to use the components :)

No support does'nt mean it does'nt work;) In fact I have'nt gone back to 1.02 since my tests, but if problems ... etc...

Geee 2 long typings in english for one day is good but exhausting, hope that helps, read you later:) .

Did

[manono]

Hi-

Here's what I (accidently) discovered to get rid of the stutters and freezing:

http://forum.doom9.org/showthread.php?s=&threadid=85433

Bottom line-a video with an average bitrate of over 7000 burned to CD-R plays smoothly, and none of the movies that used to freeze up briefly do so any more. The DVD-ROMs that these standalone DVD/MPEG-4 players use are cheap pieces of junk, and an upgrade might help matters. No one down in the hardware players forum seemed very interested, though.

Because AutoGK restricts the quants severely (and I restrict mine even more), AutoGK generated encodes are more likely to freeze up then those generated with the standard recommended settings of 2-31, where, during complex scenes the quants go up higher, and the average bitrate might not climb as much.

[richarddd]

@manono

Please give more detail (here or in the other thread) on what you did. Which parts did you switch, how did you hook it up, etc.?

[BigDid]

Originally posted by manono
[B]Hi-
Here's what I (accidently) discovered to get rid of the stutters and freezing:
http://forum.doom9.org/showthread.php?s=&threadid=85433
... and an upgrade might help
Aloha manono, neighbour from the north(Pacific) and Ia O Ra Na from the south :cool:

Nice post
The total height of my SAP is even shorter than the height of my PC-DvdRom player so not possible for me :( but I will advertise your solution whenever possible :)

Because AutoGK restricts the quants severely (and I restrict mine even more), AutoGK generated encodes are more likely to freeze up then those generated with the standard recommended settings of 2-31, where, during complex scenes the quants go up higher, and the average bitrate might not climb as much.

Thanks for your elaborate explanation , this is why I grow impatient to see Xvid 1.1 released because with the standards settings high bitrates are more compressed even if the limit is still there, just 15 to 20% away.
The real tool will be the high bitrate scenes degradation in percent so if it is available via hidden option, as I just requested from Len0x, I can encode the movie once, play it in MPC player (*1), check the bitrate, make the calculation, use the hidden option and ask for xx% more compression of high bitrate, re-encode and that's it.

Much easier than what I have done with the LOTR3: re-encoded 6 times to find the threshold and have a playable movie; actually 2: one is 2gb with audio 192vbr the other is 1/2 dvd with audio 224vbr because the original soundtrack is ac3-6ch and takes more than 600mb:( I will look into the Belite-Besweet thing and try a transcoding from 6 to 2ch but completly new in this domain so ... will see.

Anyhow thanks for the input, nice job you are doing in the forums, Mahalo & Maeva.

Did

(*1) Still searching a more elaborate tool than mpc player(with ffdshow) or the info from VDubMod to have accurate bitrates informations.

[BigDid]

Originally posted by richarddd
(1) I would have thought HDTV has more pixels than regular TV and therefore shows more movie detail.

(2) What do you mean by a 15-20% improvement?



Hi richarddd

(1) Exactly

(2) Do you still need details?

Read you later

Did

[manono]

I'm not sure exactly what you want to know, richardd. These things use standard (but poor quality) DVD-ROMs. If you're comfortable switching out hard drives and optical drives in a computer, it's the same thing. Find the screws that have to be removed to get the top off of the player. Then unplug the ribbon cable and the power cable (just like in a computer), undo whatever holds the DVD-ROM down, remove it, and replace with the new one. You have to make sure that it lines up properly, so the faceplate doesn't catch on anything when being opened and closed. As explained in the other thread, I had to raise the new one a bit with a shim inserted below it, and I used velcro to keep it from moving.

So, you might have to take off the cover to see exactly how yours is held in place, and then maybe collect some small parts to do the job (in my case, I fashioned a shim from some wood, and went to Wal-Mart to get the Velcro with the stickum on the back).

But taking the cover off will show you what has to be done with your particular model of DVD-MPEG-4 player. In principle, it's a real simple operation. Just make sure the new one lines up properly (test it out) before closing it back up. Although I mess with computers a lot, I'd never opened up one of these things before, but it was real easy.

It used to be that before watching a movie with my wife she's ask if it was going to freeze, and I'd tense up during an action scene, wondering when the stuttering was going to start. Have you seen The Bourne Supremacy, with that great car chase scene? Or some of the action scenes in I Robot? Those scenes would have killed my player before. But both movies played with no problems at all with the upgraded one. I tell you, I haven't had so much as a hiccup since I switched out the DVD-ROM. I'd hate for anyone to spend 40 bucks and not have it work for them. But you'll never know 'til you try.

[manono]

Hey BigDid-

Still searching a more elaborate tool than mpc player or the info from VDubMod to have accurate bitrates informations.

Yeah, it would be nice if there were a Bitrate Viewer for AVI. Maybe someone will make one. ffdshow is supposed to be able to do it, but I've never tried. You know, DivX, when encoding using the Home Theater Profile has had max bitrate control for some time. I tried it a few times right after I got my player and started experiencing the stutters. It works, alright, but it made the high bitrate scenes all blocky. That's the main reason I restrict my quants in the first place-to get as even a quality as possible. The idea is to use 2-pass encoding for a specific file size (one or two CDs), but with the even quality of a one-pass quality based encode. I hate that blockiness.

I forgot about the fact that some of the DVD/MPEG-4 players are real thin. Might have to use a DVD-ROM designed for a laptop, maybe.

Where the heck are you? Are you really in French Polynesia? One of those really exotic places like Tahiti, Moorea, or Bora-bora? Of course, if you really live there, it may not seem so exotic and attractive to you. Just like Hawaii is a lot of people's idea of Heaven on Earth, but like every other place, it has its good points and its bad points. The very high cost of living, for one. Just like for you, I think.

[igor_f]

:confused:
Can somebody help me?
Why, when I choose Xvid codec and define size of movie
(for example 1 Gb), I receive only about 500 Mb in final file.
Final size about 2 times smaller, doesn't metter what size I enter.
(I notesed that size usually reduced after 2-nd pass)
How can I control final size?
With using Divx I dont have this problem, but I prefer Xvid.

I was triyng all versions beginning from 1.25, but problem the same.
And, yes, I use Xvid codec included in Auto Gordian Knot installation.

Thanks for any help!

[Filbert]

Originally posted by igor_f
Can somebody help me?
Why, when I choose Xvid codec and define size of movie
(for example 1 Gb), I receive only about 500 Mb in final file.
....

Have a look at the FAQ, especially Question 6.3...
http://forum.doom9.org/showthread.php?s=&threadid=72679

[igor_f]

I thought I looked everywhere, but I defenetly miss this FAQ, or read it not carefully!

Thank you a lot!

[BigDid]

Originally posted by manono
[B]Hey BigDid-

Still searching a more elaborate tool than mpc player or the info from VDubMod to have accurate bitrates informations.
Yeah, it would be nice if there were a Bitrate Viewer for AVI. Maybe someone will make one. ffdshow is supposed to be able to do it...
Just edited my posts because I use ffdshow in mpc player to display the "instant" bitrate, not very accurate though
...but I've never tried. You know, DivX, when encoding using the Home Theater Profile has had max bitrate control for some time. I tried it a few times right after I got my player and started experiencing the stutters. It works, alright, but it made the high bitrate scenes all blocky. That's the main reason I restrict my quants in the first place-to get as even a quality as possible. The idea is to use 2-pass encoding for a specific file size (one or two CDs), but with the even quality of a one-pass quality based encode. I hate that blockiness.
Yes the profiles may be another way to do the trick, at the expense of lower quality or blockiness, dunno what position Len0x will take on the subject.
Surfed and read a lot these past hours on the subject, I should consider myself lucky because many SAP (like the ESS richarddd owns) behave much more badly than mine.
People having SAP expects the Xvid 1.1 release to cure these problems using or cumulating the following features (no order):
-Profiles like divx, yet to come
-VBV compliancy (works but is limited -within my testing- to 15/20% improvement)
-High scenes degradation, I will try to test that feature.
...One of those really exotic places like Tahiti, Moorea, or Bora-bora? Of course, if you really live there, it may not seem so exotic and attractive to you. Just like Hawaii is a lot of people's idea of Heaven on Earth, but like every other place, it has its good points and its bad points. The very high cost of living, for one. Just like for you, I think.
Exotic Tahiti? like Hawaii for sure, been there 2 times:D
Funny thing is that people here go to your place for shopping ;) everything is relative (french saying)
Completely OT, you dances may be more sensual and ours may be more sexy, can we start a comparison or a poll? just kidding, wrong forum anyhow but ... If you search hard and find that ebay 10.000$ thread or was it 100.000$ :devil:

Have a good day manono

Read you later

Did

[BigDid]

Originally posted by CeeJay.dk
For more manual control you can abort autogk after it begin the first pass and then open the *_movie.avs in Avs2avi-GUI.

Here you can insert and remove avisynth filters and tweak the encoding options as well as use a different matrix ...
Hi CeeJay.dk

Once you abort autogk how do you finish the encode? manually in Vdubmode? if so can you give more details .
Thanks

Did

[CeeJay.dk]

I use Avs2Avi .. the GUI version by Rainy :
http://www.avs2avi.org/rainy/avs2avid-1.5.zip

Rainy mentions it here :
http://forum.doom9.org/showthread.php?s=&threadid=71493&perpage=20&highlight=avs2avi&pagenumber=8

I do not resume AutoGK again .. it's purpose was simply to calculate good settings .. that you can later tweak further.

A small tutorial :

1) Open Avi2avi.exe

2) Add a new file ( File -> Add ) - There is also icon for this

3) Choose the file called "moviename"_movie.avs in the \agk_tmp\ directory as source

4) Choose an destination filename as output

5) Under Passes add a pass with XviD MPEG-4 Codec - configure its options to make the Encoding type a Twopass - 1st pass (the rest of the settings are inherited from the last time that the XviD codec was previously used .. that was when AutoGK tried to use it .. So AutoGK's settings remain.
Also click more right of encoding type and choose to discard first pass (we dont need it .. only the logfile)

5) Add another similar pass except make this a Twopass - 2nd pass

6) Now click presets and choose Save Current .. Its not necessary but next time you use Avs2Avi you can just edit this previously saved preset.

7) Click OK
You have now put a job into the queue.

To encode it simply press Encode

To edit the avisynth script highlight the job and press the script button.

Here you can add and remove filters - I'm experimenting with different noisefilters myself .. with good results.
I find that it's better to do more noisefiltering and choose a matrix that yields better compresability than to lower the resolution of the out too much if it's a movie thats hard to encode.

You can also preview the output of the filters before encoding by highlighting the job and pressing the preview button.
This is very useful .. tweaking to get the best settings would be a slow and tedious procedure if you had to encode the movie first.

You can also edit the .avs file beforehand and preview it in some mediaplayers .. I use Media Player Classic and it works perfectly.

[Taurus]

Originally posted by CeeJay.dk [[/B]

I know it's fun, to manipulate programs so they work the way ones want it.
But this is the AutoGk Development Thread. So please stay on the tracks. This thread is huge and if we clutter it more and more it gets almost unreadable.

Have a nice day

Cheers

Taurus

[BigDid]

@ Taurus - I am afraid I am at the origin of the slide? drift?
My goal is not to set AGK aside, on the contrary it is to see and/or inform Len0x about the new features of the Xvid 1.1 release useful to incorporate in AGK for more compatibility with the standalones (SAP). Quite everything here is new for me so sorry if inconvenient.:(

@ CeeJay.dk - I will try to use Avs2Avi; at the moment to see if the high scene degradation compression feature (only for 2nd pass, not available using AGK) is useful. If I have problems using Avs2Avi I will try the thread you pointed out or PM you if that's Ok.

Thanks for your time.

Did

/Edited: If there is a better thread or forum to use for the AGK+SAP+Highbitrate subject, please point it out/

[CeeJay.dk]

Originally posted by BigDid

@ CeeJay.dk - I will try to use Avs2Avi; at the moment to see if the high scene degradation compression feature (only for 2nd pass, not available using AGK) is useful. If I have problems using Avs2Avi I will try the thread you pointed out or PM you if that's Ok.

Thanks for your time.


You are welcome to PM me - I also Skype btw .. as CeeJay.dk ofcourse

[bourtzovlakas]

AutoGK 1.80 beta
- added support for external subtitles to be burnt into final AVI (TXT,SRT,SSA,SUB for text subs, IDX for vobsub subs)
(CTRL-F8 toggles subs selection)
- improved detection of changed parameters after first pass readjustments


http://img.photobucket.com/albums/v508/rey620/Temp/signs-yourock2.gif
http://img.photobucket.com/albums/v508/rey620/Temp/sarcasitc-bow.gif

:thanks:

[Taurus]

Originally posted by BigDid

@ Taurus
A few more words on the SAP topic.
Off topic too...
Like Manao said, there are so many variables to take into account.
(I like the Do-it-yourself enthusiast approach he has mentioned):D
On my old medion player I can play almost everything from Divx3/Divx*/Xvid at reasonable high resolutions and bitrates.
When you are talking about bitrate spikes: I have some old Divx clips
that have REAL beautiful spikes. The Player sometimes shutteres
and is back on his way.
As with Xvid's. In rare cases there are slowdowns and bad seeking on
high bitrate encodes.
I don't think that Xvid 1.1 vbv straight or loose curving will give a solution to this.
Cos, when people think, wow now I can raise my bitrate up a lot, suddenly SAP's memory is overflown with the constant to high bitrate.
As long as manufacturers don't take more care of the designs of their decoders and hardware, we will be lost.

Do some experiments. Play your test encodes with and without MP3 or AC3 sound. Try to mux it to video in different ways. In a short time you will be able to see the dis\advantages of your SAP.
Mostly stutters are because of wrong muxing.

Cheers

Taurus

ahem, my last words on this subject in this thread, I promise...

[len0x]

Originally posted by BigDid
inform Len0x about the new features of the Xvid 1.1 release useful to incorporate in AGK for more compatibility with the standalones (SAP).

Who can tell me - are those features available in single pass encoding with XviD 1.1? (I just don't have it installed)

[BigDid]

Originally posted by len0x
Who can tell me - are those features available in single pass encoding with XviD 1.1? (I just don't have it installed)
Hi Len0x, see here (from Koepi gui Xvid 1.1-04/11/06):

http://img131.exs.cx/img131/2868/Xvidconfig.png

1/On the left no new profiles implemented for the moment.
2/It seems nothing new in single pass (the more checkbox leads to CBR features)
3/on the right,(new) defaults for 2nd pass: loose curve scaling when not checked = strict curve,+ respect VBV buffer(nice), bottom in the curve compression item the high bitrate scenes degradation% should be working OK.

In AGK at the moment the new defaults works (as you had guessed right) but not the high bitrate% which is zeroed by AGK, I suppose before or at the beginning of the 2nd pass.

For more infos better asks the Xvid guys (me no specialist) or sharktooth who seems to have tested it quite completely.

Good luck for the next milestone

Did

/Edited: I can put more screenshots from from single pass & 1st pass later on but not at the moment/

[len0x]

As I expected this feature is rather not that useful (its bacically the same as in DivX and I don't use it either). Since 1 pass encoding cannot have this setting - it means that when predicting compressibility we can be quite far off for the source that have lots of bitrate spikes (because we predict maximum quality without taking into account rate control). One good news though - that the actual quality will be always better that we predict (since we need less bitrate with VBV) - bad news that you might have had higher resolution with the same settings at still good quality if you new that. In the worst case you might end up with undersized video because it was maxed out and AutoGK couldn't resize it because it didn't know that...

P.S. I-frame boost and reduce gets reset to 0 indeed (its not new features) because those settings ensure exact quality that we're predicting/expecting

[Taurus]

Originally posted by len0x
Who can tell me - are those features available in single pass encoding with XviD 1.1? (I just don't have it installed)

As I can see it from the Xvid 1.1(koepi)gui, the only new advantage there is "use vbv for b frames too" for single pass encodings.
To manipulate loose or straight curve scaling, it's only available in two pass mode.
Haven't tried it out.

Cheers

Taurus

Sorry, len0x you just answered when I hit he button:D

[len0x]

Originally posted by Taurus
As I can see it from the Xvid 1.1(koepi)gui, the only new advantage there is "use vbv for b frames too" for single pass encodings.


I'm puzzled - how VBV can be used in 1 pass encoding if its settings are in the second pass window...

[Taurus]

Originally posted by len0x
I'm puzzled - how VBV can be used in 1 pass encoding if its settings are in the second pass window...
Sorry typo:mad:
Sure it has to be use "vhq for b frames too"....
The only way to manipulate is under the "twopass - 2nd pass" tab.
Shame, shame on me.

Cheers

Taurus

[BigDid]

Hi Len0x, more screenshots(from Koepi gui Xvid 1.1-04/11/06):

http://img21.exs.cx/img21/1764/Xvidconfig2.png

1/On the left single pass, more checkbox.
2/On the right,the advanced options with the "use VHQ for B-frames", not enabled by default.

I will need some time to fully understand your statements (VBV) concerning the new options and AGK, I'll come back later with it, hoping you can spare some time for more explanations or confirmations.

Good luck for what AGK will be, I'm curious though :) :( :) :confused:

Did

[Sharktooth]

Ok, in the koepi's 1.1 build, VBV is based on the DivX Home Theatre profile without any restrictions other than bitrate (for example MPEG-4 tools are still selectable).
It can be used in 2 pass mode ONLY. The same applies to the curve scaling method (loose or strict).
Strict curve scaling results in less PSNR spikes (low and high) so the resulting encode min and max PSNR are more close to the average PSNR.
With loose curve scaling the codecs behaves like (but not properly the same as) the old 1.0x.
Bframes VHQ is ALWAYS a good choice for a better overall quality (slows things down a bit...).

[Taurus]

Originally posted by Sharktooth
Ok, in the koepi's 1.1 build, VBV is based on the DivX Home Theatre profile.....

Thank you, Sharktooth for this short and competent summarizing.
This is something to remember.

Cheers

Taurus

[len0x]

Originally posted by BigDid
Good luck for what AGK will be, I'm curious though :) :( :) :confused:


Do not worry - nothing that you could see will happen anyway.

[len0x]

Originally posted by Sharktooth
Bframes VHQ is ALWAYS a good choice for a better overall quality (slows things down a bit...).

Indeed a nice option, how slow is it?

[BigDid]

Originally posted by len0x
As I expected this feature is rather not that useful (its bacically the same as in DivX and I don't use it either). Since 1 pass encoding cannot have this setting - it means that when predicting compressibility we can be quite far off for the source that have lots of bitrate spikes (because we predict maximum quality without taking into account rate control).
Ok, is it 1 pass encoding only or 1 pass encoding and 1st pass of 2 pass encoding that cannot benefit from the VBV feature ?
One good news though - that the actual quality will be always better that we predict (since we need less bitrate with VBV)
Ok, the movies I have encoded with Xvid 1.1 are usually (predicted) near or very near to the expected quality
- bad news that you might have had higher resolution with the same settings at still good quality if you new that.
Can the AGK 1st pass redo take care of that resolution problem ?
In the worst case you might end up with undersized video because it was maxed out and AutoGK couldn't resize it because it didn't know that...
Uhh… maxed out how ?, the encode having the same resolution as the original (720 width to 720 width) ?
P.S. I-frame boost and reduce gets reset to 0 indeed (its not new features) because those settings ensure exact quality that we're predicting/expecting
Ok, beside the predicting, can this feature be used for less spikes when encoding for SAP ? (as an option maybe)

Lets see if I have it right :

In the past, older revs of AGK relied heavily on the comptest for the expected quality 55/60% for soft matrix below 608 width for 16/9 and 65/70% quality for sharp matrix above 608 width

Actually AGK tries to tune many variables in the 1st pass redo like
- matrix choice
- resizer choice (got sharp matrix & soft resizer together one time, been surprised)
- resolution
to get near the quality percentage wanted

Question: Is your concern to continue in that direction and, whenever easy, incorporate new features (so when not easy just let them down) ?

My concern, since discovering the highbitrate freezes in my SAP, is to ask you to incorporate as many new features as possible to help SAP playing (not so simple). In that perspective I have experimented/tested a few things and given the results in this thread. Up to now summary
- 1.1-strict curve is better than 1.02-loose curve
- 1.1-strict curve and VBV (or 1.1-loose curve & VBV -I have no tools to make the difference-) are much better than 1.02-loose curve
See datasheet and graphs here: http://forum.doom9.org/attachment.php?postid=568408 and here: http://img83.imageshack.us/my.php?loc=img83&image=Complet2.png

I have not tested the high bitrate degradation enough to say it is useful (1 time only asked –20% got near –10% compression on the High I frames – as seen in drf analyser) so maybe I asked that feature too soon, I was just re-posting what I have read in others threads or forums.

My intention in asking you to incorporate this features is not to push you to do so, just asking...
Exemple next milestone is DV/AVI input into AGK, I would have preferred the advanced/hidden options but I will wait (Joke: Hey guys, use the survey to push that milestone up :-) )

OT, puzzled by your response to what AGK will be, will try some (possible?) answers in the milestone thread

If you have some time for more details or answers I will appreciate it.
Continue the good work, everything you do is useful.

Did

/Edited: that's a long post, after reading it again and correcting some typos, I am feeling the tone/tune? of the post maybe perceived as ...hummm sharp/incisive? (still searching a better word); if so, it was not my intention/

[Taurus]

Originally posted by len0x
Indeed a nice option, how slow is it?

Just a few numbers:

Koepi Xvid 1.02 First Pass: 2.55 Minutes
Koepi Xvid 1.02 Second Pass: 3.45 Minutes 1533 kbit/s

Koepi Xvid 1.1 First Pass: 2.57 Minutes
Koepi Xvid 1.1 Second Pass: 4.02 Minutes 1537 kbit/s

Encoders set to default settings. Mpeg Matrix. 720x400 Res.

On Xvid 1.1: First pass: "Use vhq for bframes too" enabled
Second pass:"Use vhq for bframes too" enabled
Straight curve scaling, respect vbv buffer enabled.

Expected filesize ok on both versions.
If you scale the amount of time to "real life" conditions thats about 7.5% speed loss (for this encode). I don't know the behaviour on different sources or avs scipts.
My testing script was generated by Autogk. Dvd source clip, crop and resize, removegrain, that's all.
No tweaks to the encoder. VdubMod 1.5.4.1

Cheers

Taurus

[CeeJay.dk]

That doesn't sound too shabby .. in fact thats pretty good and IMO worth the extra quality :)

[richarddd]

My concern, since discovering the highbitrate freezes in my SAP, is to ask you to incorporate as many new features as possible to help SAP playing
I agree completely.

[BigDid]

Originally posted by len0x
Indeed a nice option, how slow is it?

3 Tests

Same Input: LOTR3 Vob n°4 1gb
Ouput 150 mb
No audio

Test1
AGK, Xvid 1.1
defaults: strict curve+ VBV (no VHQ)
comp test
[24/11/2004 18:52:01] Duration was: 1 minute, 28 seconds
[24/11/2004 18:52:01] Speed was: 20,68 fps.
[24/11/2004 18:52:01] Compressibility percentage is: 49,79
[24/11/2004 18:52:01] Chosen resolution is: 608x256 ( AR: 2,38 )
[24/11/2004 18:52:01] Predicted comptest value is: 55,69
[24/11/2004 18:52:01] Running first pass.
[24/11/2004 19:05:28] Duration was: 13 minutes 26 seconds
[24/11/2004 19:05:28] Speed was: 45,26 fps.
[24/11/2004 19:05:28] Expected quality of first pass size: 54,50
[24/11/2004 19:05:28] Running second pass.
[24/11/2004 19:23:37] Duration was: 18 minutes 9 seconds
[24/11/2004 19:23:37] Speed was: 33,51 fps.
Drf analyser:
Performance Caracteristics:
Macroblocks per frame: 608
The Width is multiple of 32
Kilobits per Second: 837.02
Kilobits per Frame: 33.47
Kilobits per Macroblock: 0.055
Bits per Pixel: 0.22
Frame Type Statistics :
I Frames: 1.32%
P Frames: 53.49%
B Frames: 45.19%
Frame Quality Statistics :
DRF=1&2: 2 0.0%
DRF=3: 10149 28.2%
DRF=4: 9379 26.0%
DRF=5: 5592 15.5%
DRF=6: 6800 18.9%
DRF=7: 4102 11.4%
DRF=8: 0 0.0%
DRF=9: 0 0.0%
DRF>9: 0 0.0%
KeyF/DeltaF: 1.34%
KeyDRF<4: 234
KeyDRF=4: 248
KeyDRF>4: 0
AverageKeyDRF: 3.51
MAXDRF: 7
AverageDRF: 4.58
Deviation: 1.36

Snapshot 1 B-frame + Graph here: http://img12.exs.cx/img12/3175/BframeNoVHQ.th.jpg (http://img12.exs.cx/my.php?loc=img12&image=BframeNoVHQ.jpg)
http://img3.exs.cx/img3/264/Default1_1.th.png (http://img3.exs.cx/my.php?loc=img3&image=Default1_1.png)
-----------------------------
Test2
AGK, Xvid 1.1
defaults + VHQ for B frames
comp test
[24/11/2004 19:31:12] Duration was: 1 minute, 35 seconds
[24/11/2004 19:31:12] Speed was: 19,13 fps.
[24/11/2004 19:31:12] Compressibility percentage is: 49,25
[24/11/2004 19:31:12] Chosen resolution is: 608x256 ( AR: 2,38 )
[24/11/2004 19:31:12] Predicted comptest value is: 55,09
[24/11/2004 19:31:12] Running first pass.
[24/11/2004 19:44:41] Duration was: 13 minutes 28 seconds
[24/11/2004 19:44:41] Speed was: 45,13 fps.
[24/11/2004 19:44:41] Expected quality of first pass size: 54,50
[24/11/2004 19:44:41] Running second pass.
[24/11/2004 20:04:21] Duration was: 19 minutes 40 seconds
[24/11/2004 20:04:21] Speed was: 30,93 fps.
Drf Analyser
Performance Caracteristics:
Macroblocks per frame: 608
The Width is multiple of 32
Kilobits per Second: 836.99
Kilobits per Frame: 33.47
Kilobits per Macroblock: 0.055
Bits per Pixel: 0.22
Frame Type Statistics :
I Frames: 1.32%
P Frames: 53.49%
B Frames: 45.19%
Frame Quality Statistics :
DRF=1&2: 1 0.0%
DRF=3: 9356 26.0%
DRF=4: 10170 28.2%
DRF=5: 4705 13.1%
DRF=6: 7125 19.8%
DRF=7: 4667 13.0%
DRF=8: 0 0.0%
DRF=9: 0 0.0%
DRF>9: 0 0.0%
KeyF/DeltaF: 1.34%
KeyDRF<4: 216
KeyDRF=4: 266
KeyDRF>4: 0
AverageKeyDRF: 3.55
MAXDRF: 7
AverageDRF: 4.64
Deviation: 1.38
Snapshot 1 B-frame + Graph here: http://img107.exs.cx/img107/6717/BframeWiVHQ.th.jpg (http://img107.exs.cx/my.php?loc=img107&image=BframeWiVHQ.jpg)
http://img77.exs.cx/img77/4939/DefaultWiVHQ.th.png (http://img77.exs.cx/my.php?loc=img77&image=DefaultWiVHQ.png)
-------------------------------
Test3
AGK, Xvid 1.02
defaults
comp test
[26/11/2004 06:20:43] Duration was: 1 minute, 30 seconds
[26/11/2004 06:20:43] Speed was: 20,10 fps.
[26/11/2004 06:20:43] Compressibility percentage is: 49,28
[26/11/2004 06:20:43] Chosen resolution is: 608x256 ( AR: 2,38 )
[26/11/2004 06:20:43] Predicted comptest value is: 55,12
[26/11/2004 06:20:43] Running first pass.
[26/11/2004 06:34:17] Duration was: 13 minutes 34 seconds
[26/11/2004 06:34:17] Speed was: 44,82 fps.
[26/11/2004 06:34:17] Expected quality of first pass size: 52,67
[26/11/2004 06:34:17] Running second pass.
[26/11/2004 06:52:34] Duration was: 18 minutes 16 seconds
[26/11/2004 06:52:34] Speed was: 33,28 fps.
Drf Analyser
Kilobits per Second: 836.40
Kilobits per Frame: 33.45
Kilobits per Macroblock: 0.055
Bits per Pixel: 0.22
Frame Type Statistics :
I Frames: 1.50%
P Frames: 55.13%
B Frames: 43.37%
Frame Quality Statistics :
DRF=1&2: 186 0.5%
DRF=3: 10622 29.5%
DRF=4: 10402 28.9%
DRF=5: 4977 13.8%
DRF=6: 6039 16.8%
DRF=7: 3731 10.4%
DRF=8: 0 0.0%
DRF=9: 0 0.0%
DRF>9: 0 0.0%
KeyF/DeltaF: 1.53%
KeyDRF<4: 287
KeyDRF=4: 262
KeyDRF>4: 0
AverageKeyDRF: 3.47
MAXDRF: 7
AverageDRF: 4.46
Deviation: 1.35

Snapshot 1 B-frame + Graph here: http://img90.exs.cx/img90/2889/Bframe1_02.th.jpg (http://img90.exs.cx/my.php?loc=img90&image=Bframe1_02.jpg)
http://img3.exs.cx/img3/3504/Default1_02.th.png (http://img3.exs.cx/my.php?loc=img3&image=Default1_02.png)
--------------------------------

I did not mix-up the DRF analysis (double checked), no VHQ seems marginally better than with VHQ.
/Edited: the results (DRF analyser) seem strange because Xvid 1.1(beta) is not used in a standard way by the actual AGK (1.80)/
/Edited2: 1 more test with Xvid 102, standard encode, same source, same conditions, snapshots to come/
/Edited3: Added Snapshots from DrfAnalyser Graphs/

Did

[Taurus]

Originally posted by BigDid
How did you get this thing working:

http://img123.exs.cx/img123/564/11defaults.png



http://img123.exs.cx/img123/3423/AutoGk.png

First picture is Xvid 1.1 defaults.
Second picture is AutoGk setting.
So, no use inside AutoGk.

Cheers

Taurus

[len0x]

Originally posted by BigDid
Ok, is it 1 pass encoding only or 1 pass encoding and 1st pass of 2 pass encoding that cannot benefit from the VBV feature ?
Ok, the movies I have encoded with Xvid 1.1 are usually (predicted) near or very near to the expected quality
Uhh… maxed out how ?, the encode having the same resolution as the original (720 width to 720 width) ?


You didn't grasp what is it all about :)
Maximum bitrate limit means that you have spare bits in case spike that was found is too sharp. So those bits can be used in encoding of surrounding(? I don't really know how it works exactly) frames. So basically you'll get better quality for those frames where spare bits were spent. And if you have lots of spare bits (= lots of sharp spikes) then you'll have lots of frames that have better quality that you expected/predicted. major problem arises when most of the frames are maxed out (i.e. when say compressibility is ~95-100%). You can't put more bits into maxed out frame, so you'll end up with undersized file. Having said that - it probably can only occur when you use fixed width or maximum width in autogk (or as you said you have output width 720 in auto mode already) since AutoGK aims to ~70% quality and its highly unlikely that second pass adjustments will increase it by 30%.

Moreover its improssible to say what is the real quality of the file you get (so you will never know if its just 2-3% or 20% difference to what you expect), because all adjustmets are made during second pass of XviD and it doesn't log what it was actually doing. So its impossible to detect that VBV/curves did increase compressibility quite a lot and make adjustments according to that.

Originally posted by BigDid
Ok, beside the predicting, can this feature be used for less spikes when encoding for SAP ? (as an option maybe)


yes, it is a useful feature indeed, I'm just hesitant to put it as a default option because not all standalones have this problem today (and less and less will) and because of the problems above.

This feature will be dealt with somehow for sure but only when 1.1 gets stable (and as you can set those extra options in XviD by yourself and AutoGK won't reset them - I don't really see what the problem is).

[richarddd]

@VBV Buffer

For the slow among us, to use the VBV buffer in an attempt to get better results on a SAP, just install xvid 1.1 and check "Respect VBV Buffer"? Any other settings? Also, you'd only have to check it once, as AGK doesn't reset it?

[len0x]

yes

[Sharktooth]

The VBV buffer should be safe for the everyday 2 pass encodes. IIRC the VBV limit for the DivX Home Theatre profile is about 7000kbps... wich is pretty high.
Since it works for 2 pass encodings only, 1 pass quality based encodings arent affected by VBV...
The B-Frame VHQ slowdown depends on how many consecutive b-frames are set.

[BigDid]

Originally posted by Taurus
How did you get this thing working...

Hi Taurus,

AGK is using Xvid 1.1, not me because I don't know how to use it by myself.

Concerning the parameters, you are right, just checked it. Is it serious doctor?

I mean is this abnormal behaviour affecting enough AGK so the resulting is +really+ screwed?

Joke: I already stated Len0x will not support Xvid1.1 until officially released, :D hope this statement is enough not to get sued, I am not in very good terms with my lawyer actually :mad:

Thanks

Did

/Edited:
@ taurus: do you mind if I post and quote in the xvid 1.1 thread?/

[len0x]

Originally posted by Taurus
So, no use inside AutoGk.


What does this mean?
You're getting ppl confused... (well at least BigDid who's got so nervous that started making jokes, that I don't get :) )

[BigDid]

Originally posted by len0x
What does this mean?
You're getting ppl confused... (well at least BigDid who's got so nervous that started making jokes, that I don't get :) )

Not a good day for me :( , I will edit my joke but that's no fun :D

Len0x: more to come, struggling with the quotes at the moment.

Did

[BigDid]

Originally posted by len0x
[B]You didn't grasp what is it all about :)
Yes this is why I ask. Give me some time.

This feature will be dealt with somehow for sure but only when 1.1 gets stable (and as you can set those extra options in XviD by yourself and AutoGK won't reset them - I don't really see what the problem is).
Ahem... I was talking about the high bitrate scene degradation which is (to my actual -bad- knowledge of the process) zeroed by AGK. I thought we were talking about the same thing, now not so sure; quote from one of you previous post:

" P.S. I-frame boost and reduce gets reset to 0 indeed (its not new features) because those settings ensure exact quality that we're predicting/expecting "

Thanks

Did

OT: I still dont get the following, you are welcome to come back on th subject again (now is it really a subject?) quote:
"Do not worry - nothing that you could see will happen anyway."

[len0x]

Originally posted by BigDid
Ahem... I was talking about the high bitrate scene degradation which is (to my actual -bad- knowledge of the process) zeroed by AGK. I thought we were talking about the same thing, now not so sure

Aha, but what's the problem with that? I thought VBV takes care of it anyway...

[mightyspawn]

Well ive got a DVD, Born on the fourth of july. I think it's some cheap version because it is 16:9 widescreen. But autio gordian knot says it is 4:3. When it is finished and i open the file it is widescreen, without any black borders. But the results are so terrible. Ok the original is also not that superior, but not this bad. I ripped 2x700 with 192kb audio in XVID. is pal dvd

Is this link any help for me? http://forum.doom9.org/showthread.php?s=&threadid=77780

[len0x]

@mightyspawn

lets start with :logfile:

P.S. as to your other problem - you can use File Mode if you don't need subs and just first audio track.

[mightyspawn]

[9-11-2004 14:23:28] AutoGK 1.60
[9-11-2004 14:23:28] OS: WinXP (5.1.2600).2
[9-11-2004 14:23:28] Job started.
[9-11-2004 14:23:28] Input dir: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS
[9-11-2004 14:23:28] Output file: E:\BORN_ON_THE_FOURTH_OF_JULY\4.avi
[9-11-2004 14:23:28] Audio: English
[9-11-2004 14:23:28] Subtitles: Nederlands : Normal
[9-11-2004 14:23:28] Codec: XviD
[9-11-2004 14:23:28] Target size: 1400Mb
[9-11-2004 14:23:28] Custom audio settings: VBR MP3 with average bitrate: 160Kbps
[9-11-2004 14:23:28] Started encoding.
[9-11-2004 14:23:28] Demuxing and indexing.
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_1.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_2.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_3.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_4.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_5.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_6.VOB
[9-11-2004 14:26:34] Processing file: E:\BORN_ON_THE_FOURTH_OF_JULY\VIDEO_TS\VTS_01_7.VOB
[9-11-2004 14:26:34] Source aspect ratio: 4:3
[9-11-2004 14:26:34] Source resolution: 720x576
[9-11-2004 14:26:34] Found PAL source.
[9-11-2004 14:26:34] Analyzing source.
[9-11-2004 14:31:39] Source has percentage of interlacing in motion areas: 0,33
[9-11-2004 14:31:39] Source is considered to be progressive.
[9-11-2004 14:31:39] Found 207416 frames
[9-11-2004 14:31:39] Encoding audio.
[9-11-2004 14:49:35] Indexing subtitles.
[9-11-2004 14:52:04] Compressing subtitles.
[9-11-2004 14:52:08] Audio size: 166,738,544 bytes (159.01 Mb)
[9-11-2004 14:52:08] Overhead: 8,296,704 bytes (7.91 Mb)
[9-11-2004 14:52:08] Subtitles size: 809,620 bytes (0.77 Mb)
[9-11-2004 14:52:08] Video size: 1,292,161,532 bytes (1232.30 Mb)
[9-11-2004 14:52:08] Running compressibility test.
[9-11-2004 14:52:08] Writing the following script to E:\BORN_ON_THE_FOURTH_OF_JULY\agk_tmp\4_comptest.avs
===========================================================
LoadPlugin("H:\PROGRA~1\AutoGK\DGMPGDec\DGDecode.dll")
LoadPlugin("H:\PROGRA~1\AutoGK\filters\autocrop.dll")
LoadPlugin("H:\PROGRA~1\AutoGK\filters\undot.dll")

movie = mpeg2source("E:\BORN_ON_THE_FOURTH_OF_JULY\agk_tmp\4.d2v")
cropclip = autocrop(movie,mode=0,wmultof=4,hmultof=4,samples=10,aspect=0,threshold=34,samplestartframe=0)
fixed_aspect = 1.06666666666667
c_width = width(cropclip)
c_height = round(height(cropclip) / fixed_aspect)
input_par = float(c_width)/float(c_height)
input_par = input_par > 1.4 ? input_par : (4.0/3.0)
out_width = 704
out_height = round(float(out_width) / input_par)
hmod = out_height - (floor(out_height / 16 ) * 16)
out_height = (hmod > 4) ? (out_height + (16 - hmod)) : (out_height - hmod)
new_aspect = (float(out_width) / float(out_height)) / fixed_aspect
autocrop(movie,mode=0,wmultof=4,hmultof=4,samples=10,aspect=new_aspect,threshold=34,samplestartframe=0)
Undot()
LanczosResize(out_width,out_height)
SelectRangeEvery(300,15)
===========================================================
[9-11-2004 14:58:24] Duration was: 6 minutes 15 seconds
[9-11-2004 14:58:24] Speed was: 27,61 fps.
[9-11-2004 14:58:24] Compressibility percentage is: 50,54
[9-11-2004 14:58:24] Using softer matrix.
[9-11-2004 14:58:24] Chosen resolution is: 608x256 ( AR: 2,38 )
[9-11-2004 14:58:24] Predicted comptest value is: 65,83
[9-11-2004 14:58:24] Running first pass.
[9-11-2004 14:58:24] Writing the following script to E:\BORN_ON_THE_FOURTH_OF_JULY\agk_tmp\4_movie.avs
===========================================================
LoadPlugin("H:\PROGRA~1\AutoGK\DGMPGDec\DGDecode.dll")
LoadPlugin("H:\PROGRA~1\AutoGK\filters\autocrop.dll")
LoadPlugin("H:\PROGRA~1\AutoGK\filters\undot.dll")

movie = mpeg2source("E:\BORN_ON_THE_FOURTH_OF_JULY\agk_tmp\4.d2v")
cropclip = autocrop(movie,mode=0,wmultof=4,hmultof=4,samples=10,aspect=0,threshold=34,samplestartframe=0)
fixed_aspect = 1.06666666666667
c_width = width(cropclip)
c_height = round(height(cropclip) / fixed_aspect)
input_par = float(c_width)/float(c_height)
input_par = input_par > 1.4 ? input_par : (4.0/3.0)
out_width = 608
out_height = round(float(out_width) / input_par)
hmod = out_height - (floor(out_height / 16 ) * 16)
out_height = (hmod > 4) ? (out_height + (16 - hmod)) : (out_height - hmod)
new_aspect = (float(out_width) / float(out_height)) / fixed_aspect
autocrop(movie,mode=0,wmultof=4,hmultof=4,samples=10,aspect=new_aspect,threshold=34,samplestartframe=0)
Undot()
LanczosResize(out_width,out_height)
===========================================================
[9-11-2004 15:54:37] Duration was: 56 minutes 13 seconds
[9-11-2004 15:54:37] Speed was: 61,49 fps.
[9-11-2004 15:54:37] Expected quality of first pass size: 82,70
[9-11-2004 15:54:37] Running second pass.
[9-11-2004 17:29:41] Duration was: 1 hour, 35 minutes 3 seconds
[9-11-2004 17:29:41] Speed was: 36,37 fps.
[9-11-2004 17:29:42] Splitting video into: E:\BORN_ON_THE_FOURTH_OF_JULY\4.cd1.avi
[9-11-2004 17:31:06] Splitting and compressing subtitles.
[9-11-2004 17:31:06] Executing: rundll32.exe vobsub.dll,Cutter E:\BORN_ON_THE_FOURTH_OF_JULY\4 E:\BORN_ON_THE_FOURTH_OF_JULY\4.cd1 1 0:0:0.0 01:08:17.760
[9-11-2004 17:31:08] Splitting video into: E:\BORN_ON_THE_FOURTH_OF_JULY\4.cd2.avi
[9-11-2004 17:32:46] Splitting and compressing subtitles.
[9-11-2004 17:32:46] Executing: rundll32.exe vobsub.dll,Cutter E:\BORN_ON_THE_FOURTH_OF_JULY\4 E:\BORN_ON_THE_FOURTH_OF_JULY\4.cd2 1 01:08:17.760 02:18:28.639
[9-11-2004 17:32:48] Job finished.

[len0x]

so what is so terrible about that encode?