QuEnc 0.51 (AVS->M2V/M1V Encoder)

[audioslave]

To begin with I just wanna say hi to you all. This is my first post on this forums :) ! And I wanna thank Nic for this incredible GUI/encoder. Very easy to use and with great quality encodes!

@dragongodz
when using an avi source dont use AviSource() with avisynth but DirectShowSource(). you should now find the colours are normal.
I always use *.avs (AviSynth scripts) files for my encodings so I never use AviSource(). In my test I used Mpeg2Source()... ;)

[audioslave]

@All
Comparison page updated:
http://hea.port5.com/Comparisons/Comparisons.htm
Added info on what settings I used for each encoder + the script I used.

[dragongodz]

audioslave - AviSource() is for avisynth, the load avi equivilent of Mpeg2Source(). :)
i had a quick look at your script, hmm, i would try cutting it back and adding 1 thing at a time to see where the colour change happens, which function.
before even that though try loading you .avs in MediaPlayerClassic and see if it has the same colour problem. if yes then it is definatly the script or 1 of the plugins thats the problem.

JDay - how did you do the frame comparisons ? if you search around here you will find this has happened before and sometimes it is not real. that is its is a displayed on pc is different because of hardware or drivers or overlay etc problem. thats not to say it is not real but that is why i ask how you compared, so i can try to emulate it.

Jorel - you may want to try adding/testing some filters. as an example for cartoons something like msmooth followed by msharpen. why both ? well msmooth to clean up stray blocks and rubbish and msharpen to sharpen the edges up a touch. they are quite slow and there are other similar filters of which you can find more information in the avisynth section.

Nic - is anything straightforward with libavcodec ? :D

[audioslave]

@dragongodz
i had a quick look at your script, hmm, i would try cutting it back and adding 1 thing at a time to see where the colour change happens, which function.
Yes, that would be one way of finding the source of this "problem". But on the other hand, I used the exact same script with TMPGEnc and as you can see on the screenshots the TMPGEnc sample doesn't have that problem...:confused:

[dapipa]

NIC,YOU'RE GODLY GOD,or something...
:D

i'm very pleased with your encoder,although v0.45,which i tested,was not so good at dark/red(especially DARK RED)parts of my source,but i hope v0.47'll fix it and i just can't wait to get home & try it out!

[jepkre]

Nic,

Since you're looking for suggestions on how to improve QuEnc here's my input, or perhaps not if QuEnc already does what I suggest. It seems to me that certain filters/strength and settings within libavcodec either need to be changed or aren't needed when encoding at different bitrates. For example on my encodes less than 1000kbps I've noticed that Trellis reduces picture quality by adding blocks. So maybe you could have internal settings based on bitrate(i.e., 500-800kbps uses set1 filters/strength & settings, 801-1275kbps set2 etc...). I believe this approach would improve overall quality & speed and continue to keep the gui simple(heck, even trellis & hq could be moved to advanced).

Other than that great work! For us avisynth addicts unwilling to handover "all" control it is a godsend. If at some point you feel that you are no longer able to continue development and are unwilling to release the source please find someone capable and motivated to continue with this great little encoder!!!

Thanks :)

[JDay]

The lightness in those previous screenshots may have been due to the manner that I took them. The "original" was just the avs snapshoted in vdub. for quenc, I just put the m2v into vdub. I tried the same clip with cce (4 pass, 6000 kbps) and experienced the same lightness and only moderately better quality (even though cce used twice the passes and 2000 or so more bitrate).
Here's another comparison (note: I was unable to grab exaclty the same frame in cce, pretty close, though):

AVS (http://apotheosis.tv/files/avs2.png)|QuEnc (http://apotheosis.tv/files/quenc2.png)|CCE (http://apotheosis.tv/files/CCE2.png)

Another issue, however: It says "Error: Could not open Video Codec" when you try to encode a one-frame clip at a bitrate less than 1249. You get this error on 2-frame clips at bitrates less than 947. I don't have time to do any more testing now, but this pattern may continue.

[Prodater64]

Hi: How can quenc load avs scripts? Source code are released?

[Nic]

@JDay: Well spotted again JDay. I think they're might be a bug in there...ill fix that. (@everyone: however, if you dont recieve that error, then everything is ok)

@jepkre: Thats not a bad idea. But im not sure i'll have a think. It's very common for people to throw lots of parameters at libavcodec/mencoder/ffmpeg without them really knowing what they do. I only add a parameter if it can definitely be beneficial. (e.g. I've even seen people swear by using mbd=0 ?! Which can do almost nothing but hurt the quality (although it is quicker ;) ). Although it's an extreme case, it's not uncommon)

@Prodater64:
How?
Through the medium of dance.... ;)

Release the sources?
Some forums and their users...well, one in particular...are run by people that think the GPL is b*llcr*p. That doesn't make me feel like releasing the source code because people might just rip it off, and seeing that libavcodec is LGPL, I don't have to release the source.
Of course, I have given the sourcecode to other people, dragongodz for example has it, but he knows not to give it out. And I know any changes he makes will be passed back to me, etc. So, basically anyone I can trust can get it.

-Nic

[DDogg]

OT:

Hmm, I fear the dishonorable actions by that unnamed forum may set off many of these types of reactions from folks like you and that we may all feel the weight of the lost trust you alluded to. The public upholding of GPL was what protected that trust. Anybody that attempts to suborn it is the real "b*llcr*p".

[jdobbs]

Through the medium of dance....
ROTFLMAO:D

[bullfrog]

Can someone write a short and easy guide for QuEnc?

Values and options recommended for Bitrate, GOP, KVCD matrix, HQ, Max b-frames, extreme and slow settings, ... And, if it is possible, a brief explanation.:-)

Thanks.

[Amenophis]

Originally posted by Nic
I've even seen people swear by using mbd=0 ?! Which can do almost nothing but hurt the quality (although it is quicker ;) ).

If you've gotten a worse quality with mbd=0, then you maybe haven't set any alternative comparison function, which would end up with no useful comparison function at all. imho this is called "mb_cmp" in the libav source, but as far as i know with that you can only get faster (but not noticeably better) then with mbd=2.

P.S.: With using all quality options the quality is realy good, great work! Unfortunately, there are still bitrates above 10.000 kbit/s occuring when using a high average bitrate (p.e.5000 kbit/s). Except of this, everything seems to work fine.

/edit:

@APF_Gandalf: as far as i know, one reason for these inverted macroblocks are the use of min-quant=1 AND a custom matrix.

[insertdisk]

Nic,
Thanks so much for the interlaced option - I am putting together a DVD with DV footage encoded at different bitrates and am gonna compare them to my usual encoder of choice (All the same clip). I get the bitrate spike too, (18224 at 8500 setting) but it happens only at the end of my footage. The end of the footage goes to static.
Quenc is looking really good. Thanks again.

[Affar]

@Bullfrog
You have here a simple guide for this programa ;)
http://www.doom9.hu/mpg/quenc.htm

@Nic
You can put some option to limit process priority? :)

Seeya an great work

[dragongodz]

audioslave - any chance you could try it with some other encoder aswell ? CCE and MainConcept have trial/demo versions. it would be interesting to see if the colour with them is the same as Tmpgenc aswell. who knows they could all look different. :)

So, basically anyone I can trust can get it.
now i am blushing. :o

Thats not a bad idea. But im not sure i'll have a think.
already been thinking about it and there are a couple i already want to test. its just a matter of when. but i will talk to you about them later anyway.

[Nic]

@Amenophis: I hope you don't feel my remarks on mbd=0 weren't aimed at you...Your settings do make some sense (I've seen other people that use mbd=0 with all manner of weird settings)....but mbd=0, although fast, should be avoided...if possible.

"as far as i know, one reason for these inverted macroblocks are the use of min-quant=1 AND a custom matrix...."

Do you have any more info on that? If that's the case ill turn off min-quant=1 when using custom matrices...

@Affar: "You can put some option to limit process priority?"
I knew there was something I'd forgotten...I'll add that in for the next version.

@bullfrog: I might contact Doom9 about me adding a little more to the guide (he wrote it). Just to quickly explain the settings. They aren't complicated...So if you need any help understanding any of them just ask.

Cheers,
-Nic

[Mug Funky]

(slightly OT)

So, basically anyone I can trust can get it.

that seems like a good arrangement for now, but i'd hope an LGPL or GPL might be considered in the future, simply out of principle.

as for people ripping sources... well as the fiasco with the unnamed forum shows, we'll look after you. i have faith in mob rule at least on this forum (though it definitely can get ugly).

i'd hate to see open software being undermined because of the inevitability that the world contains, and will always contain a***holes.

(back on topic)

by the way, the new rate control was just what i needed! thanks! *big kiss*

[Mozart]

talking about that unnamed forum and their unnamed matrix used in quenc: since it (the matrix) does nothing but high frequencies filtering of non-intracoded blocks(this means that it erases details from your source in order to increase the compressibility of b-frames mainly), and is a highly valued product of an "organisation", which may in the future sue you, Nic, in order to pay the royalties of their "product", I think it would be better to avoid such problems removing that hardcoded matrix from your encoder. You could maybe include a customizable matrix form instead of.

[Amenophis]

@Nic: At least when using a custom matrix and min-quant=1 in mencoder, i get something like this: http://mitglied.lycos.de/enc0404/inv_mb.jpg (all of them appear at borders, only about 1 of 200 Frames is affected).

Encoding the same clip with QuEnc, i didn't get these errors using the settings APF_Gandalf mentioned, everything was fine. Maybe there is an other reason for these errors (@APF_Gandalf: did you get the same errors like in the jpg?)

btw, these errors i got aren't affected by the trellis setting, so there might be indeed a different reason.

[Nocturno]

i'm not quite sure the bitrate control is totally fixed,

i did another test with a 200 frame clip. xvid/23.9 fps.

my script:

LoadPlugin("D:\avi2dvd\filters\mpeg2dec3.dll")
v1 = avisource("D:\test\test.avi",False)
v2 = Blankclip()
AudioDub(v1,v2)
LanczosResize(720,424,0.0,0.6)
AddBorders(0,28,0,28)
trim(0,2000)
ConvertToYV12()
#converttoyuy2() used for the CCE conversion.


all encodes at 7909 bitrate
Normal quenc 2pass came out :

59,693,679 bytes /2pass,vbr,high qual,advanced settings default.

2pass using extreme and slow settings, rest unchanged:

70,954,232 bytes

cbr high quality, extreme and slow off:

59,635,162 Bytes

CCE Encode, CBR same bitrate :

82,509,824 Bytes

that's better than .45 version, but not quite spot on yet.

thanx Nic for your hard work!

[Nocturno]

sorry it's a 2000 frame clip.. typo

[dragongodz]

no the rate control is still not 100% yet. it is getting better though. :)

[DDogg]

Originally posted my Mozart
...I think it would be better to avoid such problems removing that hardcoded matrix from your encoder. You could maybe include a customizable matrix form instead of.Mozart, nic, being the hardheaded Englishman that he is <way big grin here nic>, feels allowing custom matrices is non desirable. I would also like to ask again for this option at some time in the future =). I do find them occasionally useful, although agree totally they can be misused.

[APF_Gandalf]

Originally posted by Amenophis

@APF_Gandalf: as far as i know, one reason for these inverted macroblocks are the use of min-quant=1 AND a custom matrix. [/B]

thanks for the info, but it seems to work fine now, I just rebooted and I can't see any inverted block in my new encodes now. Maybe windows 2k3 was in a bad mood or my CPUs were overheating.
so it seems that treillis and custom matrices are safe to use together.
And the quality is really good, even at low bitrates.
It only needs support for smp systems to be perfect ;p

[audioslave]

APF_Gandalf wrote:
thanks for the info, but it seems to work fine now, I just rebooted and I can't see any inverted block in my new encodes now. Maybe windows 2k3 was in a bad mood or my CPUs were overheating.
Well, first of all, I'm glad the problem no longer exists for you, but I got inverted block on my encoding of Gothika last night :( . Don't know what's causing this...? I rebooted just before I started the encode.

I must congratulate Nic to an outstanding encoder :) . Except for the inverted block the encode is one of the best (if not )the best ;) ) I've ever done - and seen!
If only this inverted block issue will be solved there's not a question about what encoder I'm gonna use from now on :D !

BTW I did a 2 pass encode of Gothika - target was a 1 CD encode - I took the average bitrate value given by CalcuMatic (thank's Kwag ;) ), and my final muxed movie size: 794 MB :eek: :D . Bye, bye prediction!

Once again, thank you very much Nic ;) .

[dragongodz]

audioslave - yes real low bitrate encoding is a bit FUBAR at the moment. i have already messaged Nic about some observations.

[bullfrog]

Originally posted by Nic
@bullfrog: I might contact Doom9 about me adding a little more to the guide (he wrote it). Just to quickly explain the settings. They aren't complicated...So if you need any help understanding any of them just ask.

Cheers,
-Nic [/B]

For PAL KVCD and KSVCD, What are the correct settings?

- Checked notch matrix KVCD? What means "Notch"?
- GOP = 25?
- Max B-Frames: 2?
- Checked Force Closed GOP?
- Bitrate is average bitrate (with Calcumatic)?
- Use Trellin Quant?
- AVySynth script can be any script or a simple script with Mpegsource, mepg2dec3 and Resize method?


Thanks for all.

[Abond]

I have some problems with GOP structure. I have 2 max B frames and expected to have IBBPBBPBBPBB... (15 GOP). But in two encodes I made at the begining the GOP is in general IBPBPBPB... and sometimes IPBPBPB... At the middle is the same, but at the end of the file I have even IPPPPPP... or IPPPPBPBPB... GOP. I don't know if it can be fixed.
Thanks!
Edit: Forgot about settings:
MPEG2, VBR (1800), High Quality, 15 GOP size, 2 Max B-Frames, KVCD Notch,4:3.

[LigH]

In electronic circuits, a "notch filter" is a very sharp/narrow filter which is able to eliminate a very narrow band of frequencies, or a whole range of frequencies very sharply.

I don't know the KVCD technology in detail, but they use custom quantization matrices, and a notch matrix would most probably eliminate several frequency coefficienty totally by using very high coefficient quantizers (like 255) directly besides rather normal values (up to ~64).
__

Sorry, I was wrong:

1st place in the Google search resuts for "KVCD notch matrix":

http://www.kvcd.org/portal/mobile/articles.php?lng=en&pg=20

So why not searching before guessing... :rolleyes:

[Doom9]

@bullfrog: xVCD is evil if the x ain't an S.. same goes for xDVD if the x isn't a blank space. I'd strongly adwise you not to mess with the specs but to stick to them. Everything might just work fine, until you try to play your disc on a somewhat older or higher end device, and then you have no clue what is wrong because it worked on the 5 Taiwanese el-cheapo players you have at some, so it can't be the way you made the disc, can it?

And all those settings you aske for are explained for regular DVD encoding.. just read my guide.

[Amenophis]

Originally posted by Abond
I have some problems with GOP structure. I have 2 max B frames and expected to have IBBPBBPBBPBB... (15 GOP). But in two encodes I made at the begining the GOP is in general IBPBPBPB... and sometimes IPBPBPB... At the middle is the same, but at the end of the file I have even IPPPPPP... or IPPPPBPBPB... GOP. I don't know if it can be fixed.
Thanks!

It's not a bug, it's a feature ;) When using the extreme and slow settings, the encoder will try to choose a gop pattern, which will result in the best quality using the given bitrate. so your encode is better then with the pattern you wanted the encoder to use. the values 'max B Frames' and 'GOP-Lenght' are in this case the maximum values which the encoder may use. the only bug in this feature seems to be, that it will never use 2 B-Frames.

[Abond]

Ah, now is clear. But I didn't use "Extreme and Slow"...

[dragongodz]

Amenophis - actually they can happen with 1 pass, CBR, trellis off etc aswell. so not exactly a feature.

[darkluna]

Nic: First, thanks for your great program!

I have a ? in regards to using QuEnc for creating SVCDs: I'm using 0.47, which appears to have much better bitrate control; however, I have a situation where it appears that specific scenes appear to encode over the SVCD bit-rate limit, which causes stuttering during playback. I'm not sure what the MPEG stamp is, but I suspect that may be where the problem lies. Here's an example of what Pulldown shows for a particular encode:
-------------------------------
>pulldown csi.miami.2x14_quenc.m2v csi.miami.2x14_quenc.pulldown.m2v
- Enabling 2:3 pulldown
Total Frames = 62968
Total GOPS = 6030
Average bitrate = 2298688 bits/sec - (MPEG stamped @ 6900000 bits/sec)
-------------------------------
The bitrate average that I'm using is OK (2250), but the "stamp" is at 6900 bits/sec, which would appear to exceed SVCD standards. Here are the settings I'm using:

\quenc.exe -i \csim.2x14\csim.2x14_quenc.avs -o \csim.2x14\mpeg\csim.2x14_quenc.m2v -b 2250 -2 -mpeg2 -trell -aspectratio 4:3 -auto -close

Any suggestions?

[Paced]

Originally posted by darkluna
Nic: First, thanks for your great program!

I have a ? in regards to using QuEnc for creating SVCDs: I'm using 0.47, which appears to have much better bitrate control; however, I have a situation where it appears that specific scenes appear to encode over the SVCD bit-rate limit, which causes stuttering during playback. I'm not sure what the MPEG stamp is, but I suspect that may be where the problem lies. Here's an example of what Pulldown shows for a particular encode:
-------------------------------
>pulldown csi.miami.2x14_quenc.m2v csi.miami.2x14_quenc.pulldown.m2v
- Enabling 2:3 pulldown
Total Frames = 62968
Total GOPS = 6030
Average bitrate = 2298688 bits/sec - (MPEG stamped @ 6900000 bits/sec)
-------------------------------
The bitrate average that I'm using is OK (2250), but the "stamp" is at 6900 bits/sec, which would appear to exceed SVCD standards. Here are the settings I'm using:

\quenc.exe -i \csim.2x14\csim.2x14_quenc.avs -o \csim.2x14\mpeg\csim.2x14_quenc.m2v -b 2250 -2 -mpeg2 -trell -aspectratio 4:3 -auto -close

Any suggestions?

The problem is, you can't specify a maximum bitrate in QuEnc, that's why you'll always get an xSVCD output (unless you turn off VBR). I'm pretty sure QuEnc was only created for DVDs in mind, mainly due to the latest hype with jdobbs' DVD Rebuilder.

[Nic]

No, it has all MPEG-1/2 Encoding in mind. Not just DVD.

Here's the problem. libavcodec's bitrate control just doesn't work very well. You can do what a lot of mencoder/ffmpeg gui's do and set the bitrate tolerance so low that very little actual VBR happens and your mainly getting a pretty CBR stream. But that isn't good for quality...

So it ends up being a matter of tweaking. Stupidly in 0.47 CBR is broken and the max_bitrate isn't capped at any particular rate. I'll try to solve this soon.

Also a secondary problem is the bitrate can get very high at the start and/or the end of the video. This is also a bug in the rate control of libavcodec...

So that's the problem...

-Nic

ps
So close to having DC precision sorted as well...just one little bug left, but I don't know what causes it... :(

[jorel]

Nic,
is possible one little guide for command lines to QuENC and a general "how to"? thanks.

[Teegedeck]

Hi Nic,

I messed a bit with QuEnc yesterday and the day before and just want to tell you that I'm impressed. Very, very nice indeed!

The result of encoding in "extreme & slow" mode was extraordinarily detail-rich and sharp to my eyes. Unfortunately, I've tried it on a very noisy source where it amplified that effect. I'm looking forward to experimenting with it on more HQ sources from now on. Thank you, and keep it up!

[samuraijack]

I am having trouble with "Could not Open Video Codec" errors in Quenc.
Kill Bill Vol 1, R1
Ripped with DVDDecryptor ( latest version)All files
DVD Rebuilder .45
Batch encode
Quenc .47
No Trellis
2 pass, HQ, Use VBR

Seems to encounter this error when is is in the process mode, but it doesnt occur at the same place each time.

Any suggestions or anything I can do to help the process along?

Thanks for any help.
NIC, Ive heard great things about this encoder, I can t wait to see the results. Keep up the good work and thanks!

[winny]

I'm also experiencing this same error with quenc 0.47 and dvd-rebuilder 0.45, source disk is The Godfather part 3 mounted with Daemon Tools.

Quenc options:- Trellis quant, 2 Pass, Use VBR, High Quality, Extreme & Slow

Does anyone have any suggestions for troubleshooting so I can provide more information to the thread?

Thanks.

[Affar]

I've had the same problem as u when trying to compress a video with disadjusted framerate (25.002). To solve it u should add this in the avisynth script:


converttoyuy2()
convertfps(25)# Or the correct framerate
converttoyv12()

Seeya

[JDay]

I've gotten the "Could not open Video Codec" on short clips at low bitrates. It seems the shorter a clip is, the higher a bitrate is required to successfully encode. This seems to vary depending on the frame rate. A single-frame 29.97 fps clip needs a bitrate of 1561 to encode. A 23.976 fps clip needs 1249. (same frame, just different fps!)

This problem is common with DVD-RB, as it will often try to encode single-frame clips at very low bitrates (if dynamically assign cell bitrates is on). You can get around this by manually encoding the troublesome clips at the required bitrate, or turning "Dynamical assign..." off (assuming the average rate is greater than the needed one).

[winny]

Excellent, thanks for the suggestions. Will give them a try and let you know how I get on.

Many thanks.

[Nic]

Hi,

I don't know what would be causing this Can't open video codec problem. Unless the data from the AVS File was or settings were...strange. Hmmm. Can anyone re-create this problem. And if so can they tell me how to?

I've almost finished a new release (max bitrate (doesn't work 100%, still get a spike at the start but generally better), scene detection on/off, etc)...but i'd like to fix this first.

Thanks.

-Nic

[JDay]

I've gotten the "Could not load Video Codec" error with different sources (both mpeg2source and directshowsource). You should be able to recreate this with a script like this:

(load whatever video, don't think it matters)
lanczosresize(720,480)
changefps(29.97) #this should have a min rate of 1561.
trim(1,1) #or whatever frame, it hasn't made a difference in my testing
converttoyv12()

I'm at work right now, so I can't do any more testing but that should help recreate it.

Edit: This was commented in the avs but I moved it here because it broke the tables: you can use changefps(23.976) instead and the min rate will be 1249. Or you can just leave it at whatever the source is, but the results may vary (I've only tested on ntsc sources, and therefore, haven't tried other framerates or resolutions yet...).

[winny]

Hi Nic,

I've since tried another source disk, REM - Perfect Square, and I think JDay is onto something here.

The AVS script you suspected is as follows:-

LoadPlugin("J:\dvdr\decodefix100\MPEG2Dec3dg.dll")
mpeg2source("J:\DVDRB\D2VAVS\V02.D2V")
trim(211692,211716)
ConvertToYV12()
AudioDub(BlankClip())

But the corresponding entry in rebuilder.ecl shows a very low bitrate (i am using the default option for dynamic bitrate allocation):-

vbr_brate_avg=180

This disk has more examples of these single cells with the same entries in the ECL above.

Hope this helps. Let me know if you need further info.

[samuraijack]

I tried Kill Bill again last night with the dynamically Assign set to Off. It went off without a hitch, although it did take a quite a while. I took a look at the resulting VOBs this morning and they look very crisp. I think I will burn these to a DVD and compare them to the version I managed with CCE. I have to admit, I am VERY impressed at how nice this looks. Nice Job, Nic!

[Nic]

http://nic.dnsalias.com/QuEnc.exe

Changes:


Now has DC Precision for MPEG-2
You can turn off/on Scene Detection. Having it off will cause more regular GOP patterns
Can set Max. Bitrate. You still get a bitrate spike either at the start or end, but generally it will be obeyed.


The "Can't open video codec" error hasn't been fixed yet...I need to do more bug testing. But that bug should be rare....I need more info on it.

Cheers,
-Nic

[JDay]

More info on the "Video Codec" error: It seems to pop up when the bitrate is set very low for a clip. I tried clicking all the settings on and off, and I think I may have found the culprit. I turned off "2 Pass" and do not get the error now. The other settings seem to have no effect, although the framerate of the avs seems to affect the min bitrate that can be used with 2pass on. When I turned off 2pass, it would even encode with bitrate set to 0 (although the output was WAY oversized for 0kbps ;)).

Edit: Also, what exactly is "DC precision"? I googled it but got results on voltage meters or something.

Edit the second: Also, also, I noticed that vbr and hq are still missing from the commandline. Could you add those, pretty please?