Auto Gordian Knot: current version 2.55

[len0x]

Version 1.5 was never released yet. Can you point me to the download site ?

[Shuichiro]

http://www.chip.de/downloads/c_downloads_11784819.html

I hope you can say anything about the situation.

My Error... 1.25 was offered for download. They just put it as a new download. Sorry ^^ Anyway hope to see 1.5 soon ^^

[Carraway]

Originally posted by len0x
I'm doing slight modification to AutoGK these days: hidden option for enabling sharp matrix for small sources, increasing comp test frames for short sources (to make it more precise). What do you thiunk should be limit of frames - 2000, 3000, 5000?

I'm dense. When you say "limit" -- do you mean the limit of frames in the comp test, or the cut-off point for what would be a "short source"?

Originally posted by len0x
Also I switched soft matrix to Javor's 1CD matrix.

I'm personally thrilled about this. :)

[len0x]

Originally posted by Carraway
I'm dense. When you say "limit" -- do you mean the limit of frames in the comp test, or the cut-off point for what would be a "short source"?


Its lower limit for number of frames examined. So either 5% of the source or 2000 frames whichever is greater.

[jeremymacmull]

btw has anyone looked at tomshardware.com as in the testing for the sempron processor they use AUTOGK with divx 5.2 and xvid 1.01

i thought autogk did not support divx 5.2 (they use 1.48beta btw)

just interesting that a hardware site would use your program to benchmark

and even more interesting as supposedly autogk does not support 5.2 yet !!

JEREMY

PS on the other hand if autogk does support 5.2 i apologise but its still interesting to see sites use the proggie for benchmarks

[len0x]

Originally posted by jeremymacmull
btw has anyone looked at tomshardware.com as in the testing for the sempron processor they use AUTOGK with divx 5.2 and xvid 1.01


They do that for a while (at least 4-5 months)

Originally posted by jeremymacmull
i thought autogk did not support divx 5.2 (they use 1.48beta btw)


Correct, but if you don't mess with DivX settings after you installed it then AutoGK will work fine. The only parameter I added for 5.2 support for version 1.50 was quantanization type (so if you set up it as MPEG2 then you'll screw your encode but version 1.50 will not allow that).

[jeremymacmull]

cheers very informative

not looked at tomshardware in abit so i sort of missed the 4-5 months bit

JEREMY

[len0x]

Originally posted by Carraway
I'm personally thrilled about this. :)

I wonder if there is a better matrix than HVS best for high bitrate encodes?

[QNX]

len0x

i've uploaded sample to

http://eoninfo.kiev.ua/temp/the.matrix.1080i.hdtv.hdl.sample.avi

but it be avaliable for 2 days

i have a strip at the bottom, same as on other movies and cause of this all xvid movies also have that strip :(

[len0x]

Originally posted by QNX
i've uploaded sample to


First, I get "Bandwidth Limit Exceeded" when trying to download. Second, resulting avi is kinda useless to me - I need original source TS (or demuxed m2v).

[QNX]

this is TS, just rename it to ".TS"

it will be avaliable after 31 of july - size - 20mb

but it's only 14sec

[len0x]

Originally posted by QNX
this is TS, just rename it to ".TS"


lol, that was confusing :)

Originally posted by QNX
it will be avaliable after 31 of july - size - 20mb

ok, cheers. remind me please if I forget...

[jeremymacmull]

i really think autogk is a brill product it makes converting ds9 episodes soo much easier.

lenox can u give us a hint as to what u would like if possible to add to autogk in the future

(no timescale or pressure just an idea of what direction youd like autogk to go as i think the app is fairly feature complete anyway!)

JEREMY

[len0x]

Originally posted by jeremymacmull
lenox can u give us a hint as to what u would like if possible to add to autogk in the future


I don't really have huge plans. I'd like to release a stable version as soon as I have more or less good support for 50/60fps sources (although this means writing new analysis routine and I'm being really lazy these days, so due to the lack of sources and ppl who actually can benefit from this I might postpone this). After that there is only two big things in my mind: consolidate all hidden option into one file and possible add more customization to it (there will be lots of flaming about it I expect :) ) plus support for DV sources (or even generic dshow sources support). Support for standalone files might lead to an addition of audio option where you can specify an audio file to add to the video (but I really don't want to do that at the moment).

[jeremymacmull]

hmm for me personally not so sure about the 50/60 fps sources

cant see many people using that !?


but the consolidating all hidden stuff into one file and making it easier to customise i can see the use of that definately.

brill product cheers for the good work

JEREMY

[Alwayscamp]

:Lenux
Anyway - can you cut out very small part of TS using ProjectX(basically demux a part of video to m2v file) and make it available to me please? After we determined what the source is I might need a bigger sample for analysis...
I keep on seeing a lot of these request on this board for people to send files and i just thought i'd help with my two sense. You Send It .com (http://s2.yousendit.com/) is a great web site that lets you send pretty big files. If I remember correctly the last time I checked it was 1GB. Now that i think should be quite enough for for most files only downside is that it stays on there website for a 7 day period only.

[len0x]

Originally posted by Alwayscamp
You Send It .com (http://s2.yousendit.com/) is a great web site that lets you send pretty big files. If I remember correctly the last time I checked it was 1GB.

fantastic! very useful, thanks.

[QNX]

len0x

time to download file on page 128 :D

[ukb007]

AGK makes the .d2v file with ForceFilm OFF in DGIndex; and that should have a line in the file saying the FPS is 29970. Right ?

I expect AGK changes that line after working on the file. What other changes are made in that file?

Regards.

[len0x]

Originally posted by ukb007
AGK makes the .d2v file with ForceFilm OFF in DGIndex; and that should have a line in the file saying the FPS is 29970. Right ?


correct

Originally posted by ukb007
I expect AGK changes that line after working on the file. What other changes are made in that file?


It might change it if FF has to be on. It also changes field operation line in that case.

[len0x]

Originally posted by QNX

time to download file on page 128 :D

Ok, I did. Its 30fps telecined sources which is pure FILM @ 24fps. AutoGK should have no problems restoring 24fps. At maximum resolution you need to have a top-end PC to play it smoothly though...

[len0x]

A couple of bugs fixed, a couple of new hidden options introduced, full support for divx 5.2 and new soft matrix (Javor's) as promised (if you have suggestions for sharp matrix instead of HVS_Best please do suggest them :) )

[QNX]

Originally posted by len0x
Ok, I did. Its 30fps telecined sources which is pure FILM @ 24fps. AutoGK should have no problems restoring 24fps. At maximum resolution you need to have a top-end PC to play it smoothly though...
i think i have good enough.. :P

try to encode it.. do you have a black strip at the bottom?

[QNX]

i'm talking about such strip

http://images.people.overclockers.ru/8082.jpg

this was encoded from HDTV like that sample.. and it has that crap at the end... dunno what to do :(

[len0x]

Originally posted by QNX
this was encoded from HDTV like that sample.. and it has that crap at the end... dunno what to do :(

In the sample you provided it does have a small black bar at the bottom ~50% of the time. So some scenes have it and some don't. Obviously there is no automatic solution for that as autocrop will always fail (nomatter what the settings are) to crop full video frames...

P.S. what is this sample that "brakes", i.e. shuttering as I understand? (coz I see no problem playing resulting avi at 720*xxx).

[Taurus]

version 1.50 beta is up
@len0x You are fast as lightning.
Thanks for the new beta.
It includes almost everything discussed the last weeks.
Hmmh, one thing. Why did you change to Jawor's matrix?
Thats right: Jawor not Javor as you spelled it. Is this another person?:confused: :D
I was using HVS matrices - good, better, best for a long time now.
Did many a testing on other matrices. Always switched back to the chosen (older)ones.
What are the benefits on this for AutoGk? I did quite a few tests on Jawors matrice, but really can't tell the difference to HVS-good.
Depending on source sometimes Hvs-good,sometimes Jawor's is better.
So maybe a little explanation?

My sharp matrix suggestion: Stay with HVS_best.
Sure there are many matrices which taylor out more details (pump up the volume).
But to my opinion HVS_best is still the best choice comparing filesize and details without to much compromising.

But to live is to learn;)

Cheers

Taurus
P.S.: Jawor's matrix is really good on 1CD rips.

[QNX]

P.S. what is this sample that "brakes", i.e. shuttering as I understand? (coz I see no problem playing resulting avi at 720*xxx).
what? :D

i mean that it's skipping frames of playing them not so fast

[len0x]

Originally posted by Taurus
Thats right: Jawor not Javor as you spelled it.


:D I edited it :)

Originally posted by Taurus
What are the benefits on this for AutoGk? I did quite a few tests on Jawors matrice, but really can't tell the difference to HVS-good.
Depending on source sometimes Hvs-good,sometimes Jawor's is better.
So maybe a little explanation?


AutoGK was using HVS_better, not _good matrix actually. I can't see difference myself, but Jawor's matrix quite a bit more compressible (up to 10% on average). So now 1CD rips should have large resolutions.

Originally posted by Taurus
My sharp matrix suggestion: Stay with HVS_best.
Sure there are many matrices which taylor out more details (pump up the volume).


I really wanna have a matrix which resolves more details without huge increase in size. This only possible on high bitrates otherwise we get artifacts. May be I can even make AutoGK work with three matrixes. Atm I'm looking at "EQM V2" and "Soulhunters v6" this friday blind test should be over, so I'm curious to see results...

[len0x]

Originally posted by QNX
what? :D

you said you wasted 10 hour on encoding and movie "brakes"... and it was 60fps TS... remember? is this about provided sample ?

[QNX]

Originally posted by len0x
you said you wasted 10 hour on encoding and movie "brakes"... and it was 60fps TS... remember? is this about provided sample ?

yes.. it's about such sample, but the movie was - FF :D

[Sharktooth]

Originally posted by len0x
Atm I'm looking at "EQM V2" and "Soulhunters v6" this friday blind test should be over, so I'm curious to see results...
EQM V3 is coming... :D
However i'm curious to see how V1 and V2 performed. Those are both good matrices.

[PlazzTT]

Website down now?

Fatal error: Call to undefined function: message_die() in /home/len0x/public_html/web/db/db.php on line 88

[thop]

Would it be possible to have AutoGk accept a time range? No fancy graphical stuff, just like "00:10:10 - 00:20:54".

[Codesmith]

I am someone who had no problem running GK, however I did find that I had to skim over the guides if a month lasped between rips.

Now with AutoGK I am getting the exact same quality with few seconds effort!

So far I have only run across one DVD which AutoGK couldn't handle.

I am actually dreading having to fire up regular GK to attempt to encode the problem DVD. AutoGK has me spoiled.

[jsquare]

Just discovered AutoGK, coming back to Divx after a few years hiatus, using back then Vidomi and recently trying Nero Recode 2 plus a few others applications, I found AutoGK to be the the easiest use and best achieved looking AVIs.

But I have a little problem with the latest beta(1.50), with stable version 1.25 I was able to open the VIDEO_TS folder on my mounted decrypted ISOs and the movie info will pop-up, now with version 1.50 it's asking to rip the mounted DVD with Decrypter to IFO files, somehting that seems as a step backwards, so I went back to v1.25.

Is there any chance that this option will be included on newer version?

[thop]

jsquare: http://forum.doom9.org/showthread.php?s=&threadid=77941

[jsquare]

Originally posted by thop
jsquare: http://forum.doom9.org/showthread.php?s=&threadid=77941

That's a real drawback, guess I'll keep using v1.25.

[blade12]

Originally posted by len0x
how long exactly ? (if manually demuxed)

I noticed that the demux step is now 40 mins, cheers.

However, I have had instances where demuxing for a 2 hour movie takes roughly 1 hour.

Since no-one else seems to be having this problem, it might be a sign for me to get a better laptop (or HDD at least) :)

[bwalerud]

Have tried to connect twice recently without success

[killingspree]

Originally posted by jsquare
That's a real drawback, guess I'll keep using v1.25.

how do you consider this a drawback - if you're ripping the dvd in ifo mode you save disk space and you do not have to mount the dvd again! the ifo parsing method has shown to have some serious drawbacks, so honestly i do consider it to be the better - or easier solution!

also the new versions have been improved a lot concerning encoding quality and behind the scenes calculations so actually using v 1.25 (although stable) would be much more of a drawback imho!

also pls note that encoding to divx from files on your hd should actually be noticeably faster than encoding from an iso file as you save a considerable amount of cpu time as it doesn't require you to read from the iso file on the fly!

of course you're free to use any version you want :)

kr
steVe

edit: just to confirm: can't access the agknot site either!

@len0x: wouldn't it be a good idea to close this thread and start a new gknot dev thread (2nd edition)? in this thread it is extremely hard to find any information you might need - reading through 130 pages isn't really a quick way to retrieve any information ;)

[len0x]

Originally posted by bwalerud
Have tried to connect twice recently without success

Looks like its up now. It was probably down the whole night (GMT time). Bad timing :)

[len0x]

Originally posted by killingspree
@len0x: wouldn't it be a good idea to close this thread and start a new gknot dev thread (2nd edition)? in this thread it is extremely hard to find any information you might need - reading through 130 pages isn't really a quick way to retrieve any information ;)

I think we all agree that this thread is no longer suitable for finding info :) I treat it as free chat place (i.e. talks like "website is down" would rather not have separate threads). This is main development thread and starting another one will result in another long thread as well. All comprehensive topics should ideally have their own threads. No one expects users to read the whole thread - they are feel free to post questions answered 50 pages ago again :)

[len0x]

Originally posted by thop
Would it be possible to have AutoGk accept a time range? No fancy graphical stuff, just like "00:10:10 - 00:20:54".

You mean trim video? Have a look at FAQ Q6.6

[len0x]

Originally posted by QNX
yes.. it's about such sample, but the movie was - FF :D

I dunno about FF but in my case it was properly IVTC-ed...

[jsquare]

Originally posted by killingspree
how do you consider this a drawback - if you're ripping the dvd in ifo mode you save disk space and you do not have to mount the dvd again! the ifo parsing method has shown to have some serious drawbacks, so honestly i do consider it to be the better - or easier solution!

also the new versions have been improved a lot concerning encoding quality and behind the scenes calculations so actually using v 1.25 (although stable) would be much more of a drawback imho!

also pls note that encoding to divx from files on your hd should actually be noticeably faster than encoding from an iso file as you save a considerable amount of cpu time as it doesn't require you to read from the iso file on the fly!

of course you're free to use any version you want :)

kr
steVe


Got it, now I understand why it's necessary after having audio sync delays with one movie last night using v1.25.

Maybe an implementation of vStrip could be included in the future, just like DVD2SVCD, that way AutoGK will become a one-click DVD2AVI solution.

By the way, I tried v1.5 on another system and managed to fit Raiders of the Lost Ark in 1-CD at 720x320, it took almost 4hrs but came out really good for an 1h:55m movie in full resolution and the audio was superb.

[Carraway]

Originally posted by len0x
(if you have suggestions for sharp matrix instead of HVS_Best please do suggest them :) )

I'm by no means an expert on this, but I've found Didée's SixOfNine-HVS is pretty great for high bitrate encodes.

Originally posted by len0x
Sort of autosize mode, right? I though about it a lot and the only reasonable solution would be:

- add option to adjust size when fixed or minimal width selected
- this option adds 700Mb increments if good quality is not reached
- this option won't change audio type
- this option will only work for 1/2CDs original target size and make in the end either 1, 2 or 3 CDs

If ppl are happy with this scenario then I can implement it.

I wanted to bring this particular idea back because I think it's a good one.

My speciality is stupid ideas, so here's one: why not make an autosize feature that is ridiculously basic? i.e. if, say, the rip was between 15-30 minutes use 175MB, 30-60 minutes use 350MB, 60-100 use 1CD, 100-200 2CD's, > 200 would be 3CD's, etc. Something like that, where the number of frames is used to determine the inital size decision. That would be used as the basis for the autosize decision, and then after a compressibility test, it could be determined if the size should be stepped up/down one level. Also, if the video length is less than 15 minutes, I guess AutoGK could encode by average kbps or quality-based (although then the risk would be run of having a file greater 175MB even though it's less than 15 minutes). P.S., the minute values I listed are pretty arbitrary, ther eare probably better values to use.

I apologize if it's a horrible idea or if I'm missing some huge reason as to why it simply wouldn't work. Length is just usually what I initially base my size decision on, and then if it turns out to be incredibly compressible or not, I'll adjust it accordingly. It seems to work reasonably well. Now it would just be automated. :)

Even if this is a bad idea I still think having an autosize would be a nice feature, since it would add that one extra step of automation. If anyone has any better ideas (which, let's be honest, is pretty likely), I say post 'em.

[len0x]

Originally posted by Carraway
I'm by no means an expert on this, but I've found Didée's SixOfNine-HVS is pretty great for high bitrate encodes.


how does it compare to HVS_best in compressibility ?


Originally posted by Carraway
why not make an autosize feature that is ridiculously basic? i.e. if, say, the rip was between 15-30 minutes use 175MB, 30-60 minutes use 350MB, 60-100 use 1CD, 100-200 2CD's, > 200 would be 3CD's, etc.

There are still some choices about audio and initial matrix (also now that I have HDTV sources input - resolution-wise its difficult to predict sizes like you described). To be honest I think that now with with XviD 1.0.1 and more compressible soft matrix desired size can be achived almost all the time with decent resolution even with auto-width mode. So there are less and less reasons for having auto-size mode...

[Carraway]

Originally posted by len0x
how does it compare to HVS_best in compressibility ?

I did a very quick test at quant 2 with otherwise default xvid settings.

01-6of9.avi -- 8,996kb
01-hvsbest.avi -- 5,958kb

02-6of9.avi -- 72,136kb
02-hvsbest.avi -- 51,320kb

03-6of9.avi -- 16,692kb
03-hvsbest.avi -- 13,034kb

So, yeah, a whole lot less compressible than HVS_best. Definitely makes for a great looking image, but sadly very hard to compress.

I am ashamed.

Originally posted by len0x
There are still some choices about audio and initial matrix (also now that I have HDTV sources input - resolution-wise its difficult to predict sizes like you described). To be honest I think that now with with XviD 1.0.1 and more compressible soft matrix desired size can be achived almost all the time with decent resolution even with auto-width mode. So there are less and less reasons for having auto-size mode...

That's understandable. The HDTV sources completely slipped my mind, but that would definitely make any resolution-based judgement on autosizing pretty much impossible.

Today's score: 0 for 2! :)

[((( atom )))]

@len0x,

back again..

after lots of comparing encodes on 2600 ansi projectors i really would like to ask again for any (hidden) option to set the comp. threshold to ones liking.

my experience is, that when watching encodes on either tft displays or - worse - very good projectors, the nature of the digital displaying technology with its perfect but hard pixels boost visibility of artifacs far over the visible threshold compared to crt monitors or tvs.

is it only me seing that?

i compared several autogk encodes with gk encodes and lowering the resolution simply brings more detail@less artifacts up until i'd say the good old 80% comp value.

so please don't get me wrong here, my motives are of course being lazy (and so wanting to use autogk more again) as well as believing that everybody could have nicer encodes.

[thop]

(((atom))), you're right. I'd try ffdshows noise at a low value. Improved the experience for me on my TFT.