AutoQMatEnc - a new free MPEG2 encoder

[SAPSTAR]

Same error problem again,except now it still gives the error even if i put the proggie in the Matrix folder like the work around before.

g:)
As I said before ...could you give more info !!! Just saying there is an error is not really helpful !!! What is printed in the DOS window for example?!?!?

[BadServo]

Well, I ran a series of tests to compare the quality of the current AQE with HC. Unfortunately, I used a bitrate too high for the source material, making visual comparisons very difficult, I intend to perform the tests again with a much lower bitrate, in effort to introduce more artifacts, making side by side comparison a bit easier.

However, in a effort to show users of tranditional transcoders the error of their ways, I'll post this one comparitive shot, in which DVDShrink butchers a relatively simple image.

Also included are a number of stats reguarding the processing time and output size. Check it out here. (http://badservo.com/external_links/AQESMHICOMP.htm)

[alfixdvd]

@BadServo

In your test : AQE-2 Pass w/QMatOp.

w/QMatOp it's with or it's without ?

[BadServo]

With QMatOp

[borgraf]

I'm looking forward to testing this encoder, and i'm all for people doing comparisons and showing them. It's always interesting to see what others come up with. I have a comment here though:
However, in a effort to show using of tranditional transcoders teh error of their waysSettings: ... DVDShrink-Deep w/No Adaptive CorrectionIn all fairness then portray "the user of the traditional transcoder DVDShrink" in a fair way and use AEC in your tests. It'd actually be interesting to see what difference it made :)

[BadServo]

Fair enough, and I didn't mean to sound derogatory towards those who choose traditional transcoders as their meathod of choice. Merely a cheeky joke on my part. Truth be told, I'm quite impressed with the relative quality that CloneDVD produces by default with such a high rate of speed.

As far as AEC goes, this is largely a metter of personal taste. Some perfer smoothness, some perfer sharpness, and each is used to better effect on different material. I chose to leave it disabled in my tests to give a more basline comparison. However, if your sincerely curious, I'll toss in a couple Shrink shots with AEC in my next comparison.

When I use Shrink, I generally go with Sharp. Any objection?

[SAPSTAR]

@BadServo

In your test : AQE-2 Pass w/QMatOp.

w/QMatOp it's with or it's without ?
Of course, we are talking about the included engine, not the standalone version !!!! CCEAQM=1 in REBUILDER.INI to enable it.....

[alfixdvd]

Of course, we are talking about the included engine, not the standalone version !!!! CCEAQM=1 in REBUILDER.INI to enable it.....

I agree.

In BadServo's test: Encoding Time: HC-230min, AQE-174min

HC is 0.14 ?

AQE is the last 0.22a or prior version ?

[gilligan2]

As I said before ...could you give more info !!! Just saying there is an error is not really helpful !!! What is printed in the DOS window for example?!?!?


As i said also,the same error as it was giving before which was

"ERROR: File V0100000(...) shows completed, but no .m2v file exists. Aborted"

Sorry if i wasn't clear.

I have it set up the same way that it worked with the work around of putting the proggie in the DVD Rebuilder matrix folder and it didn't work. I tried to put the new one in the same folder and it still didn't work like the old way did.
I also tried it just in its own folder first like normal and that didn't work either.

Thanks.

g:)

[SAPSTAR]

I agree.

In BadServo's test: Encoding Time: HC-230min, AQE-174min

HC is 0.14 ?

AQE is the last 0.22a or prior version ?
it's the 0.22a2 a prior version of the 0.22a....(for beta testers only), but the speed is the same than 0.22a.

[SAPSTAR]

As i said also,the same error as it was giving before which was

"ERROR: File V0100000(...) shows completed, but no .m2v file exists. Aborted"

Sorry if i wasn't clear.

I have it set up the same way that it worked with the work around of putting the proggie in the DVD Rebuilder matrix folder and it didn't work. I tried to put the new one in the same folder and it still didn't work like the old way did.
I also tried it just in its own folder first like normal and that didn't work either.

Thanks.

g:)
Hi ! I meant the log in the AQE window.......with the log coming from DVDRB I can't do anything !

[gilligan2]

Their is no AQE log in the window because it never starts.

g:)

[borgraf]

When I use Shrink, I generally go with Sharp. Any objection?Nope. The default setting sounds like the right choice for this :)

[SAPSTAR]

Their is no AQE log in the window because it never starts.

g:)
If there is no window starting check your DVDRB settings !!! There is ALWAYS a dos window opening if DVDRB is pointing to the application....
[EDIT] Anyway, I'm to release a version with an auto-installer to get rid of this kind of issue....

[gilligan2]

OK,thanks for the help. I got it up and running !!


g:)

[SAPSTAR]

Given all errors reported about the dirs and install pb, I release ealier than expected the latest version :
0.23b - 2005/06/21 :
- Pass 1 auto restart in case of cancelation in the next passes.
- Setup program with the addition of the correct PATH for DCTune2.0 in the PATH variable
- Tweaked the rate control curve + new internal mode for ultra low bitrates (<1500)
(Not fully tested - waiting for results from BadServo on this part)
- Corrected a bug when the dirs contained spaces...stupid one this one !!!

Thanks to FishMan0919, BadServo, Carpo and danpos for all their tests !!!

[Carpo]

git - a new rls and after i started a 2 dvd batch encode ;D

[Carpo]

well after running Dr Who Vol 1 thru autoqmatenc - was really supprised with the results

total time 12 hours (ish) but.....

it was 133 mins in length (3 eps) with
UnDot().Deen() avisynth filters
passes 3 (it needed high compression (58%)) vaf + 2 passes
Jawor's 1CD Quantization Matrix (if there is better filter that i should be using feel free to post and suggest one)

Other Info:

- AVS Filters are enabled.
- "CCE Adaptive Quantizer Matrices" is enabled.
- "Movie and Menus Only" mode is enabled.
- VTS_02: 3,611,703 sectors.
-- Scanning and writing .D2V file
-- Processed 200,801 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 55.7%
- Overall Bitrate : 3,974Kbs
- Space for Video : 3,896,618KB

now in total the dvd was 4.32GB and the picture is really good - much better than i thought (sorry sapstar) although it is slower than most (which did put me off at first) the finished result is well worth the extra time.

Now im no expert on the subject but i would strongly advise giving this encoder a look over :)

[SAPSTAR]

....
UnDot().Deen() avisynth filters and
Jawor's 1CD Quantization Matrix (if there is better filter that i should be using feel free to post and suggest one)
...
About the filters...I'm using BlindPP(cpu2="xxxxxx") ....just a matter of taste I suppose...
About the speed, I'm trying to find a way to increase it...not so easy given the fact that in a 2 passes mode, there is a first pass just for QMatOp so 3 passes in fact.....I may try another strategy soon.....

[BadServo]

Now im no expert on the subject but i would strongly advise giving this encoder a look over :)

Good to hear you had such good results. I've found that with bitrates over 3000, quality is nearly identical to HC Encoder (which I think delivers generally better quality than CCE) and in some cases perhaps a bit better. At bitrates in the 2000-3000 range, it seems to be hit or miss between HC and AQE as to which is better. Depends largely on the source I would presume. On bitrates lower than 2000, HC seems to have an advantage, as more blocking appears in high-motion scenes with AQE. But this is something SAPSTAR is working tirelessly on.

Thus far, disc size has been very good. I've yet to have an oversized disc with AQE. However, occational undersizing may occur. It is interesting to note however, that in my personal tests, the undersized material produced by AQE was identical in perceptual quality to the HC results that utilize the entire disc.

I agree that SAPSTAR has made remarkable progress in a short time and has developed a VERY capable encoder. I strongly encourage anyone who hasn't tried it to do so. Them thar optimized Matrices sure help a might.

[SAPSTAR]

On bitrates lower than 2000, HC seems to have an advantage, as more blocking appears in high-motion scenes with AQE. But this is something SAPSTAR is working tirelessly on....
I would say between 1500-2000 :) under 1500 It has to be tested ! ;)
[EDIT] I ran some tests at 1000bps and on halfD1...up to now I would say hard to make a difference between HC and AQE...I would say maybe better with AQE...but I may be biased ! ;)

[MaDxCrEaM]

Is the download site down? I can't connect, tried both my computers with no luck.

[BadServo]

The site does appear to be down at the moment. However, I spoke with SAPSTAR today, and a new version with significant improvments is on the way. If we're lucky, it may be ready be tommarrow.

The only versions I have handy at the moment are the interim test versions, or I'd post it myself.

[MaDxCrEaM]

thnks for info BadServo

[Carpo]

latest on the site was 23b - i could post that if required

[archaeo]

Sapstar:
I decided to go ahead and give AQE a try on a challenging source, a 196 minute (older) film with an average bitrate of around 2280 and compression at 54%. Everything was set up properly, and AQE went ahead and went to work. However, about 30 minutes into the process, I noticed that the DOS progress window was hung up at this particular spot (see attached) and message was indicating a buffer underflow.

the log is below:

[21:07:49] One Click encoding activated...
-----------------
[21:07:49] Phase I, PREPARATION started.
- CCE SP 2.70.2.0 encoder selected.
- "CCE Adaptive Quantizer Matrices" is enabled.
- VTS_01: 3,729,964 sectors.
-- Scanning and writing .D2V file
-- Processed 301,198 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 54.1%
- Overall Bitrate : 2,281Kbs
- Space for Video : 3,497,256KB
- HIGH/LOW/TYPICAL Bitrates: 3,643/500/2,281 Kbs
[21:17:02] Phase I, PREPARATION completed in 10 minutes.
[21:17:02] Phase II ENCODING started
- Creating M2V for VTS_01 segment 0
- Creating M2V for VTS_01 segment 1
- Creating M2V for VTS_01 segment 2
-- Encoding resulted in no .M2V, retrying.
- Creating M2V for VTS_01 segment 2
Aborted.

Not sure what did this, but I aborted..Was this an error based on what you see in the DOS screen?

any ideas on how to solve this? The program looks promising.

[BadServo]

Buffer Overflow messages in the DOS box are not uncommon and are nothing to worry about. As for why M2V creation failed, I'm not sure. Thus far, I've only have this occur once, and it ended up being some things I had changed in RB-Opt that caused the problems. Perhaps your doing some extra pre-processing that's conflicting.

I'm sure Sapstar can shed some more light on this.

[Carpo]

i have had this twice with dr who vol 1 and vol 2 now, first time was cus of dir name (that seemed to be the case) other poss reason was that it was pal and interlaced - or it could have been becuase i had remove ifo structure protection turned on in dvd-rb

altho i have backed up true lies (ntsc interlaced) with no issues :mad: - so im at a loss to figure this one out

[Axlemar]

I didn't see it mentioned in this thread, but the program is now hosted at http://www.badservo.com/autoqmatenc/

Also, when I use the program I get a message about qmat_shift being larger than 21 and that possible buffer overflows can occur and then I get messages about buffer underflows being @ specific locations. Is that normal? Also how would you go about using the animation switch like in qmatop or change the tolerance or time (if you still want people messing with those settings).

[Carpo]

Also, when I use the program I get a message about qmat_shift being larger than 21 and that possible buffer overflows can occur and then I get messages about buffer underflows being @ specific locations. Is that normal?

yes they are normal and are ok as fishman has posted to me - as i asked the same question they are fixed in the 2nd pass

[Axlemar]

I finished an encode with the program (which took 14 hours) on an animated series I owned. The results came out jumbled with no indication of the original picture, just green and multi-color lines. The source was interlaced and the program detected alternate lines. I used 64-bit idct, 4 passes, and I am pretty sure I set it up properly. I am using the dgdecode that comes with rebuilder and the newest avisynth as well as the newest pro rebuilder.

[Fishman0919]

I used 64-bit idct.

IEEE 64-bit does have a known error in it and 64-bit Floating Point doesn't but that is not the problem.

and the newest avisynth.

Avisynth 2.5.6 Beta3 [050505] ???, if so bad idea... try 2.55

4 passes.

3 passes will give you the same quality as 4 passes and takes less time... the only benny to 4 passes might be better sizing on hard to encode stuff

As for the Green and Multi-Color lines....dunno... never had that happen with AQE

[Carpo]

what you dont want are buffer under runs which i have been getting and sapstar is looking into it :)

[Axlemar]

It was avisynth. I was using the newest version instead of the stable and for some reason it works with CCE SP and not AutoQMatEnc.

[danpos]

At Others MPEG1/2 Encoder forum (http://forum.doom9.org/showpost.php?p=678139&postcount=66)

CYA!

[archaeo]

Just wanted to update my last post where I hit a 'buffer underflow' notice on the DOS screen- apparently I aborted the encode w/out having to. When I ran the source again and let it alone, it went through that segment after some time. So it was not hung up as I suspected, but was just busy processing.

[Fishman0919]

Just wanted to update my last post where I hit a 'buffer underflow' notice on the DOS screen- apparently I aborted the encode w/out having to. When I ran the source again and let it alone, it went through that segment after some time. So it was not hung up as I suspected, but was just busy processing.

Sorry archaeo I didn't see your post earlier, yes if 'buffer underflow' error comes up the encoder continues ... sometimes it does look like the encoder is hanging... If I was paying more to what I was reading I would have told you this sooner..... Sorry :o

[archaeo]

no problem. thanks for the help :)

[Axlemar]

I think I have noticed something interesting. On the same animated series I backed up (now this I got it working due to originally having the wrong avisynth for this program) I got undersized results by ~250 megs doing 4 passes. During the encoding I noticed that after the buffer overflow messages that there were sometimes buffer underflow message that listed specific section of the video. The underflow didn't always happen, but when it did the expected file size was quite a bit higher than the actual file size. Also, has anybody else had a good sized encode with an animated disc using an idct other than the default? Also, using the newest beta avisynth gave me messed up video results and a long encode time, but the size was right on.

[SAPSTAR]

OK...As I said, I just wanted to share a bit of work of mine...that's becoming too complicated, and the intentions of certain persons not so clear (I'm not to make any money with my encoder nor anybody else). So I choose to stop everything here !!! I'm fed up with that licensing issue http://forum.doom9.org/showthread.php?p=678489#post678489 , it was supposed to be fun, it's not....so I'll keep my encoder and future mods for myself only. For info, I was to release the OPV compliant version.....

[Carpo]

:mad: :mad: :mad: :mad: :mad: :mad: :mad: :mad: :mad:

[Krizzz989]

Why don't you just do it xvid style, by releasing the source code (as proof-of-concept?), and we can compile it ourselves?

[buzzqw]

please SAPSTAR

don't be sad/angry

continue your work, it's a good program and many of us will benefit from using it on our backup

:)

best regards

BHH

[SAPSTAR]

please SAPSTAR

don't be sad/angry

continue your work, it's a good program and many of us will benefit from using it on our backup

:)

best regards

BHH
Don't worry I'm thinking about a way to come back "legally"...most probably by using a libavcodec.dll instead of linking ot it directly, the only sad stuff is I won't be able to modify the motion estimation procedure as I was planning to do in order to increase the speed...but the rest should be the same, as I used up to now a regular libavacodec but I overrided the rate control through the rc_override feature.....

[winny]

I hope you can work this out to everyone's satisfaction, it sounds like a worthwhile project.

Best wishes.

[danpos]

Don't worry I'm thinking about a way to come back "legally"...most probably by using a libavcodec.dll instead of linking ot it directly

Cool, man !! :cool:

:thanks:

[Carpo]

should the beta testers be on standby then :sly: :D

[Axlemar]

The quality of this encoder was really good and it takes a lot of the guess work out of finding optimal settings. Please don't let anyone discourage your brillant idea for an encoding process.

[beanbrain]

Please don't let anyone discourage your brillant idea for an encoding process.

It's not about discouraging, but all about encouraging respect for the license of the software his own encoder uses.

Xvid, Qenc, libavcodec, Linux, etc., and all the projects forked from them probably wouldn't be here if the people involved in those projects just wanted to have "fun", but didn't respect the license and release the code they developed because it was “too complicated”.

The GPL, LGPL, etc. is the glue that holds the open source community together. Without it, the likes of Microsoft and others would be free to profit off all the hard work put into the great number of open source projects without having to give back a single cent. This would collapse the community. There’s a reason Microsoft called the license a virus – it’s because it meant “hands off” to them, and they didn’t like it.

If Sapstar doesn’t like the GPL/LGPL that’s his own choice – but he still can’t use GPLd code in his own project without following the license agreement, or getting permission otherwise from the copyright holder of the code he’s using.

Even though Sapstar released his work for free, and I’m sure it never entered his mind that he was doing anything wrong, he still needs to respect the wishes of those that freely gave him access to their code via the GPL.

That aside, it's a great project and I hope he continues -- but this time with more reverence for the other authors of the code he's using.

[winny]

A cursory glance through the other forums shows your point has been made clear and hammered home in several threads already.

I hope everyone gets behind him (and any other developers for that matter) to make the implementations as successful as possible.

Hopefully everyone will benefit in the end. :)