QuEnc 0.51 (AVS->M2V/M1V Encoder)

[Amnon82]

@Nic: Can You post all commandline options?

add I found a feature You can add: CPU Priority

Is it possible to add a Pause feature?

BTW: Libavcodec after CVS 14.03.04 produces bad quality.

[dragongodz]

Matthaeus - "incorrect VBV size (36kB). It should be 23kB for VCD & 112kB for SVCD/DVD."
Nic - "Are you sure it's only 36kb ?"

yes i am curious about where the 23kb came from aswell. i have seen vbv as low as 20 for vcd but also higher. example is Tmpgenc uses 40kb for its vcd templates and MainConcept encoder uses 43kb and they are vcd compliant.


Nic - sorry i havent had a chance to get back to you but real world is interfering at the moment. :(
will try to get some testing done by the end of the weekend.

[Nic]

@avalon:
It isn't done using a Commandline so it's hard to post them, but the ones used are pretty obvious from QuEnc...apart from HQ does not equal the commandline param -hq it equals -mbd=2 which seems better to me.

Yup I can add pause and processor priority.

"Libavcodec after CVS 14.03.04 produces bad quality."
Do you know what caused it? Ill check, was it just bad quality in MPEG-2/1 or was it to all codecs (and hence perhaps a change to the motion est/comp code?)

@dgz: sorry for not replying to your mails, read them and taken the info on board, may well send you the code if you get some free time and you can tinker with it if you like, ill try and put out ReJig as well this weekend....

-Nic

[tuco]

I noticed that when I go in the command prompt and type quenc /? it doesn't show all the options that can be changed, namely Use VBR and High Quality. I am comparing with mencoder since I have been getting excellent results with but the using batch files can be a pain and I haven't found a gui for it that was suited to my needs. Your previous versions would cause the video to jump every second or two which I think is because of the b-frames. I am testing your current version as I type to see if it does the same. Yup, if I have b-frames enabled, I get the jumpy video while making a VCD. I don't get this with mencoder which is odd since both your program and mencoder use the same libavcodec. Is anyone else getting this problem?

Edit: I still ge the same problem without b-frames....Here is my script

AVISource("C:\Capture\PerryMason (2004-03-15 at 12-58-15).avi").ColorYUV(levels="PC->TV")

Trim(102198,103996)
Converttoyv12()
SimpleResize(352,240)

FluxSmooth(7,7)

Edit...Again: Well reading the two posts above me, now I know why. :)

[dragongodz]

Nic - hahaha hows that for posting at the same time ?
ye i should have a little time free sunday and also monday.

[Nic]

@tuco: If there isnt a commandline param for it, just turn it off/on then quit and it will be saved in the registry so that next time you do it from the commandline it will be done. What are you playing it in to produce the skipping, ill look at how mencoder sets it's b-frame settings...
If you believe it's because of the VBV then it should be fixed in the current version...

@vmesquita: I've made an option for the upcoming 0.5 that turns on using the KVCD matrixs for encoding, however, the quality it produces doesn't seem very good with that notch matrix...is there anything im missing?

-Nic

[yaz]

@nic
still i'd like to set max/min br by myself if possible. 9Mbps is ok for dvd but pretty high for svcd. & what about 'trellising' ? or is it defaulted ?
btw, u were right. i (quenc:-) was a bit careless with the br before.
thx
y

[dragongodz]

"still i'd like to set max/min br by myself if possible. 9Mbps is ok for dvd but pretty high for svcd"
well compliant vcd isnt a problem since its cbr and can be done already. hopwever svcd being vbr but limited max bitrate needs at the least a definition set, maybe as a selection/switch.

i am sure these sort of things will come in time. lets not forget its only at version 0.4. :)

[yaz]

@dragongodz
:-))) yeah, i know ... :-)))

[Paced]

Originally posted by Nic

@vmesquita: I've made an option for the upcoming 0.5 that turns on using the KVCD matrixs for encoding, however, the quality it produces doesn't seem very good with that notch matrix...is there anything im missing?

-Nic

If you intend on having support for KVCD, you might want to allow a (open) GOP size over 18 (currently QuEnc doesn't allow you to go over this) - KVCD users are known to go higher (I think between 24 and 30). Also, in order to get a better quality KVCD (if you didn't know already), try encoding in MPEG-1, use the KVCD notch matrix (which you're already doing), and use the optimal AVS script that was specifically made for KVCD encoding. Oh yeah, don't forget to lower your resolution to one of these: 544/480/352 x 576(480 NTSC).

[tuco]

Originally posted by Nic
@tuco: If there isnt a commandline param for it, just turn it off/on then quit and it will be saved in the registry so that next time you do it from the commandline it will be done. What are you playing it in to produce the skipping, ill look at how mencoder sets it's b-frame settings...
If you believe it's because of the VBV then it should be fixed in the current version...

@vmesquita: I've made an option for the upcoming 0.5 that turns on using the KVCD matrixs for encoding, however, the quality it produces doesn't seem very good with that notch matrix...is there anything im missing?

-Nic

I am using mplayer classic to play my files. It is more like a distorted freezing effect every second or so that makes the video look like it is jumping up and down by a few pixels. You get this sometimes with mencoder at very low bitrates, less than 700. I will wait for the next version. Take your time, your are doing a great job :)

[Amnon82]

@nic: Pics are better to discribe it:

Look here:

Encoded with libavcodec 14.03.04 (http://rj-elektronik.de/avideo/lastmencoder/quality14.03.04-20.55.jpg)

Encoded with libavcodec 17.03.04 (http://rj-elektronik.de/avideo/lastmencoder/quality17.03.04-09.30.jpg)

[gizmau]

-custom matrices
-selectable maximum bitrate
-batch encoding
-'shutdowen after encoding'-option
-bitrate-limitation of end credits

keep up the good work!!

[Nocturno]

Hi Nic,

i saw 0.4 had Slightly changed bitrate control

so i decided to run the same test.. it came out a few bytes bigger, but still way under size.
any ideas about that (same stuff i posted earlier).

[Paced]

Originally posted by Nocturno
Hi Nic,

i saw 0.4 had

so i decided to run the same test.. it came out a few bytes bigger, but still way under size.
any ideas about that (same stuff i posted earlier).

Is this movie quite 'dark' ? I've noticed that QuEnc has bitrate allocation problems with 'darker' movies (which results in an under-sized output). Also, are you doing 1-pass encoding? If so, try 2 passes, it gives much better accuracy (although, for me, darker movies still turn out under-sized even with 2 passes). One last thing, make sure the video you're testing QuEnc with is greater than 5/6 minutes in length, anything lower and you're bound to get an under-sized result.

[Amenophis]

i've tested your program too, with the following results...

- the quantizer doesn't get below ~3.0, even when using a high bitrate (and the encoder doesn't even reach the given avg bitrate). I-Frames look really good (that's what i expect when aiming 8000 kbit/s), but B-Frames in high motion scenes unfortunately not. seems as they are getting a higher quantizer. have you set 'input->b_quant_offset' to any high value? i would suggest something like -0.2 here. also 'input->b_quant_factor' should be something like 1.0...1.2 (that means b-frames will get 0%...20% more compressed than P-Frames. Additionally in mencoder there is an option to let the encoder decide how much b-frames would be the best for every gop. It's called vb_strategy, when activated it decides for every gop which pattern will be the best, imho a must-have when using mencoder (lowers avg. quant about 0.5 when used for dvd->dvdr). imho the option 'input->b_frame_strategy=1' will activate this mode (if it isn't affected by the comment above). additionally 'input->qblur=0.0' and 'input->qmin=1' should result in better quality,

- the quantizer curve seems to be very constant, so a better rate control algotithm could be useful. with 'input->rc_eq' you can pass a *char, mencoder uses something like 'tex^qComp' (with qComp=0.5 or something like that, and tex is the texture quality (maybe you've to calculate this out of itex and ptex saved in the logfile)).

- for better quality (and slower encoding ;)) the following options should be useful too:
input->dct_algo=6 //uses FAAN Algo for DCT
input->mb_qmin=1 //a single mb can be encoded with quant=1
input->lmin=1 //like the one above needed when using qmin=1
input->quantizer_noise_shaping=true //increases quality, but trellis is needed

- RcOverride, as well as masking and me-prepass seems to be interesting as well, but i haven't done much testing regarding these options.

[Nocturno]

Is this movie quite 'dark' ?
no not really.. and i'm doing a 2 pass (see my first post about this.)

[dragongodz]

Nocturno - with CCE was the output set to elemtary stream or program stream ? if program then load Tmpgenc and go to mpeg tools. go to de-multiplex and load the mpeg. double click on the video stream(you should see at least a private stream aswell) and save. this will be the raw m2v which is what QuEnc is outputting.

Amenophis -
"option to let the encoder decide how much b-frames would be the best for every gop."
ah but is that vcd/svcd/dvd compliant ?
"masking and me-prepass seems to be interesting as well, but i haven't done much testing regarding these options."
i did actually try without prepass using ffvfw some time ago and found the quality of my limited tests where very close in general. so it really needs a bunch of people testing and posting their results for something like that to become a real issue.
"the quantizer doesn't get below ~3.0, even when using a high bitrate"
hmm well i did a test with low bitrate(3000kbps) ata dvd resolution and some I frames did get down to quant 2 according to bitrate viewer. i will have to try a high bitrate test aswell and see what that does.

[Nic]

@Amenophis: Thanks for the tips. There are so many options its been hard to know which is best. FFMPEG's defaults certainly arent so ive been surfing the net finding better ones for mencoder, hopefully the next release will be more turned. (I think b-frame quant factor is set to 2.0 at present, which sounds too high then (it's the ffmpeg default))

@Nocturno: Thanks for the feedback, if you get chance, try a 2-pass and see what results you get. Thanks.

-Nic

[Razorblade2000]

In case you wanna build in the shutdown function:
On http://www.robvanderwoude.com/shutdown.html there are the different commandlines you can use to shutdown Win9x/2k/xp/nt

[dragongodz]

ok quick test done with 8500kbps results. I and P frames min quant was 2 and B frames was 4. clearly a problem there. I and P frames should be a min of 1 and B frames min of 2. you should find this is the norm for mpeg encoders.

"there are the different commandlines you can use to shutdown Win9x/2k/xp/nt"
thanks Razorblade2000 but Nic and i talked about Rejig shutting down different OS internally some time ago and he already knows how. so no need for commandline tools.

Nic - DCT precision selection needs to be added to options since 8 for vcd,9 for svcd and 10 for dvd etc. currently just does 8.

[Amenophis]

@Nic: if you need good settings for mencoder, i've done several Backups following the Big3-Way (with using Mencoder instead of CCE), and the results were even a little bit better than the CCE results with using libav options like this:

vcodec=mpeg2video:vrc_buf_size=1835:vfdct=6:keyint=15:vmax_b_frames=2:vrc_override=110200,115248,1000:vlelim=-4:vcelim=-7:trell:mbd=0:aspect=16/9:vb_strategy=1:vratetol=1835:cbp:vbitrate=4250:vrc_maxrate=8497:vqmin=1:mbqmin=1:lmin=1:vb_qfactor=1.25:vi_qfactor=0.833333:vqblur=0.0:vpass=1(pass2:=2)
@Dragongodz:

"ah but is that vcd/svcd/dvd compliant ?"

imho yes. of course only when using proper keyint and max_b_frames settings. but changing the GOP pattern within an encode shouldn't be a problem, as long each pattern is valid.

"DCT precision selection needs to be added to options since 8 for vcd,9 for svcd and 10 for dvd etc. currently just does 8."

i don't know if that is already implemented in the libav, mencoder too hasn't such an option. at least DCT prec.=8 is also valid for svcd/dvd and the quality is still very good.

[Nic]

@Razorblade2000: Thanks for the tip, but the best code for closing down windows was in Nandub, so ill use what that taught me :)

@Amenophis: Thanks alot for that. Ill add those in as pretty much default. Ive tried most of them out in testing and they do seem to help.

@dgz: Precision, as far as I know, isn't changeable, ill look at the code, maybe I can add it to libavcodec.

-Nic

[Nocturno]

@dragongodz - i used Es so nothing to demultiplex

@Nic. i did do a 2-pass

like i said i used a bitrate of 4684 on a movie of 62 minutes 23 secs

so it should come out about 2gb right?

it came out : 1,493,066,198. 2-pass
my CCE ES encode: 2,101,766,688


i'll try another conversion this afternoon.

[dragongodz]

"Precision, as far as I know, isn't changeable, ill look at the code, maybe I can add it to libavcodec."
ye i just had a look at the ffmpeg options and its not listed just selection of different algorithems.

"imho yes. of course only when using proper keyint and max_b_frames settings. but changing the GOP pattern within an encode shouldn't be a problem, as long each pattern is valid."
ah but what is a valid pattern ? a setting that can make a GOP IBBBB etc for vcd/svcd especially is suspect. look at kvcd, yes lots of modern players will play it but that in no way makes it in spec or required to be playable on future players. so thats why i believe that an option for standard type vcd/svcd/dvd should follow the normal GOP pattern only. since Nic is also allowing the ability to encode kvcd etc then max_b_frames could be used for free form mpeg encoding of course. what i am talking about is for if people want to set encoding to vcd/svcd/dvd standard that should play in all their dvd players. infact ffmpeg does have an option for setting target type(vcd/svcd/dvd) already where format options are automatically set.

Nocturno - the problem is probably related to the min quants bug that i posted earlier. that is I and P frames min quant of 2 and B frames of 4. once thats fixed to what they should be then that may fix the problem(lets hope :) ).

[Amenophis]

@Dragongodz:

Libav won't create a GOP like 'IBBBB...', because i've set max_b_frames=2. So it will only greate GOPs like 'IBBPBBP...', 'IBPBPBPBP..', 'IPPPPPPP...' or a mix of these patterns. The GOPs won't get longer than set with 'keyint'.

@Nocturno:

I'm sure that's because of the quantizer. With mencoder and (constant) quant=2 the codec needs about avg. 4000-5000 kbit/s. Currently we have here only avg. quant>3, so the encoder won't get above ~3000-4000 kbit/s (avg). but imho that's not a major problem, it can be easily fixed like mentioned above

[dragongodz]

"GOPs like 'IBBPBBP...', 'IBPBPBPBP..', 'IPPPPPPP...' or a mix of these patterns"
yes the IBBB was just an example. IPPPPP(as example) is just as suspect. so max_b_frames as an option for free form mpeg encoding i dont have a problem with. however for anyone wanting to create dead standard vcd/svcd/dvd this should be disabled. thats why i also said maybe have options for setting standard vcd/svcd/dvd encoding which would use normal GOP pattern etc etc etc. you will find that CCE, Tmpgenc, Mainconcept etc etc etc all have such selectable profiles/templates.

[Nocturno]

ok thanx guy's.

did another test on an 1:16:13 hour long movie and it too came up way short..
at a bitrate of 7909

aiming for 4,2 gb it came out 1,6 gb.

i'll just wait for you guys to decide about the quantizer :)

[Amenophis]

Originally posted by dragongodz
yes the IBBB was just an example. IPPPPP(as example) is just as suspect. so max_b_frames as an option for free form mpeg encoding i dont have a problem with. however for anyone wanting to create dead standard vcd/svcd/dvd this should be disabled.

IPPPP... and even IIII... will result in an standard dvd, i found nothing in the dvd specs which forbits these settings. so every dvd-compatible standalone player will play this stream. You can create such GOPs also with CCE and Tmpeg, and even Scenarist doesn't output any warning.

P.S.: The feature we are talking about is called 'vb_strategy' and not 'max_b_frames'. the last one is always needed for dvd streams.

[dragongodz]

"IPPPP... and even IIII... will result in an standard dvd, i found nothing in the dvd specs which forbits these settings. so every dvd-compatible standalone player will play this stream."
sorry but i dont have th $5000(U.S. dollars) to buy or desire to sign non disclosure agreement for the dvd specs. but from what i understand yes dvd players should be able to play them. though i did read somewhere of a person getting some juttering with IPPPP on a self created dvd while IBBPBB was fine, effected only 1 of his dvd players. ok so thats really the players problem but it illustrates the need for this to be selectable rather than automatically on.

"You can create such GOPs also with CCE and Tmpeg, and even Scenarist doesn't output any warning."
and you missed the point that they are not what you get when you select the preset setting or load preset template which is what i was talking about. so if Nic adds presets for vcd/svcd/dvd why should it be different than all other encoders ?

"P.S.: The feature we are talking about is called 'vb_strategy' and not 'max_b_frames'. the last one is always needed for dvd streams."
thats called a typo while busy trying to juggle doing multiple things at once. :)

also note i said vcd and svcd aswell and not just dvd. take kvcd, most dvd players will play such encoded video. it is however out of spec and in no way is mandatory to be supported. giving people the option to encode different ways is fine but if you have a standard setting/option then it should be as conservative as possible to ensure proper playback on all players.

well i am starting to feel i am banging my head against a brick wall here so i will not bother to comment on that any more. whatever Nic decides to do is his decision and should be respected as such.

[crusher]

I've been unable to encode because I keep getting the error "AVS file is not outputting (Use ConvertToYV12() at end of script." The script already has (Use ConvertToYV12() as the last line. Any ideas?

[jorel]

hy Nic and all!

yesterday i had little problems using cce
and after change some posts in cce forum with Kedirekin,
he found the problems and now everything is working fine,
but for me, cce is complicated to do the adjusts, i'm a simple user.

when i was waiting his answers, i see this thread and
was (and is) a big surprise cos i never had read about QuENC before.

then, today soon, i download QuENC to compare the results cos
i had encoded the "toy story" with cce and using the same
source and script, i did a new encode with QuENC.

was needed to remove the line ConvertToYUY2()
from the script to work with QuENC.

QuENC with default adjusts, changing only the aspect radio to 4:3 !
(and changing the colors of the gui, very nice feature too!)

final sizes:
QuENC = 1,49Gb ---> very cool!
CCE266= 1,66Gb (using Q=20,mpeg2(es one-pass vbr))

comparing quality doing fast test in powerdvd
just the same, very nice.
then i change the scenes many times to see more details:
the QuENC encode is pulsing the image every second !

anyone had seamless problem?

please, hints are welcome, QuENC is amazing and easy to use.
thanks in advance!

@ Nic :
thanks for QuENC !

:)

[djan]

Hi,

Tested QuEnc 0.41 using ReBuilder with an average bitrate of 3.198kbs. Sorry to say but I didn't get a good quality. And compared to CCE 2-pass, this one was much better for the same bitrate. Maybe I'm doing something wrong but I get a lot like flies flying around all sprites/objects, I don't know how you call this. Understanding it's in an early development state, I'm sure to see some quality improvements later in the development. I don't forget it's free ! Keep up your great work Nic.

[human75]

Create AVS script DoITFAST4U.Tested QuEnc 0.41 an average bitrate of VBR 4000. I didn't get a good quality.CCE is good on VBR 4000 but QuEnc 0.41 bad quality.what is wrong?Thanks for ideas.

[djan]

Same for me as I explained in my last post. It seems like QuEnc is not able to handle well low bitrates. But don't forget it's in early development state. We can see a great free tool further.

[tickey]

Originally posted by jorel

the QuENC encode is pulsing the image every second !

anyone had seamless problem?

Yes, I get "the alive" feature as well :D (pulsing vid)
2 bad I can't turn it off ;)

[flloyd]

Originally posted by jorel
the QuENC encode is pulsing the image every second!

Ditto here as well.

[Nic]

Thanks for the feedback guys and sorry for the delay in a new release, all this feedback helps loads so keep it coming.

I get the pulsing bug here, but (I dont think) not at home, almost wondering if it's a P4 SSE2 thing...Does anyone that doesnt have a P4 get the pulsing problem?

New update soon, I hope,
-Nic

ps
Ok it's not the SSE2, if anyone else wants to test easily, make a avs script like this:
Version()+Version()+Version()
BicubicResize(720,576)
ConvertToYV12()

And then encode, you'll see the pulsing then.

EDIT:
Praise and the winning prize goto: Amenophis :)
It's the standard iDCT algo that causes the pulsing, if I use idct=6 then no more pulsing :D
(ok idct=6 shouldn't work, but it does, must be defaulting to something else, any which way, problem solved)
EDIT2: Yup, there appears to be a bug in the mmx implementation of simple idct in libavcodec...ill do more testing, see if I can fix it. For now if you use the libmpeg2/xvid style idct, no more pulsing :D
Edit3: Ok, there appears to be a problem no matter what mmx idct I use, only if I use the standard C one is it ok. Speed isn't impacted too much. Even happens with ffmpeg.exe so it's not QuEnc's fault.

[jorel]

"....it's not QuEnc's fault."
oh yes,i had conviction about that!
:)

hey Nic,
you edit "too much" the last post,now i'm confused! (kidding but real confused)
:p

what i have to do ? (using athlon xp 2000+)

thanks !
:)

[Nic]

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

* This version uses normal, no mmx, SimpleiDCT and no longer shows any sign of the pulsing bug.
* Also has some encoder options that Amenophis suggested.

-Nic

[unplugged]

(haven't read the whole thread)

Nic, do you think to further incorporate MPEG2Dec3(dg) routines to support direct MPEG source decoding?

[Nic]

(actually i'd probably use the libmpeg2 to do direct decoding...or ffmpeg's own mpeg-2 decoding to do m2v/vob/mpg->m2v)

(edit: BTW, just tried the kvcd matrix, and with this bug fixed, the matrix works real well, so the kvcd notch matrix be in the next version of QuEnc)

[jorel]

great news Nic!
:D

after download QuENC 0.42 and
using same source and script posted that i compare with cce encode
but with trim(0,2000) ~01:23 seconds only to do a sample:

no more "pulses", very clean !
:)

now i will encode the full movie (cartoon) again!

thanks you for this great tool!

:)

[Nic]

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

* Now has a fixed simple mmx idct which doesnt cause the pulsating...(so no lose of speed hopefully and no pulsing)

* Now can select the KVCD Notch Matrix (also has commandline option to do so)

-Nic

[kwag]

Originally posted by Nic

* Now can select the KVCD Notch Matrix (also has commandline option to do so)
-Nic
Hi Nic,

Just tested it on a small clip.
Looking really nice ;)

-kwag

[HarryM]

@Nic: Great, great program... :D

Do you plan to add a support for "AVS to audio(mp2)" encoding?

"Pulsation" is smaller in v0.43, but still exist... :(

[sidders]

Hi

This is my script from DIF4U:

LoadPlugin("C:\Program Files\DoItFast4U\mpeg2dec3.dll")
Mpeg2Source("VTS__01_P01.16~9_1.d2v",idct=0)
ResampleAudio(44100)
ConvertToYV12()

I get the error AVS file is not outputing. Use ConvertToYV12()at end of script.

As u can see, that is already there. Am I doing something wrong?

Thanks

[Nic]

@HarryM: It's not ;) honestly if you see pulsing now, then it's due to something else. It's not the same pulsing that was there originally. Try it with that script I posted and you'll see it's gone.
Maybe I have I-Frame and P/B-Frame quants factors set too different, please post the settings you used.

@sidders: First don't PM me with stuff you've already posted. 2nd, I dont know what's wrong. Sometimes it can throw up that error if something goes wrong in Avisynth, have you looked at the AVS in VirtualDub or Media Player to see how it runs, looks?

-Nic

[robw]

Hi Nic

Thanks so much for sharing your encoder. Please consider posting updates to the encoder with a version number

e.g. http://nic.dnsalias.com/QuEnc_v043.exe

[Nocturno]

Hi nic

Also has some encoder options that Amenophis suggested

do these have to do with the problem i mentioned earlier..
i did another CBR test of 40 seconds at a high bitrate (7900) and came up with 50% less filesize than the same encode in CCE.