battle of the CBR/VBR/CQ modes

[adam]

It makes me laugh that you can't analyze the results correctly

I can analyze your results perfectly, the problem is they have absolutely no relavance to well...anything. Come on man you encoded one sample with mpeg1 and one with mpeg2 yet you are not attempting to test the differences between the two mpeg standards. How can these results show anything about the encoders that produced them? If you wanted to test TMPGenc's CQ quality versus its 2-pass vbr encoding would you use mpeg1 for one test and mpeg2 for the other? What is this supposed to prove? And this is just the most obvious inconsistency in your test, I already mentioned most of the others.

Adam, you always come up with your famous "subjective" word. Is CCE or a sponsor paying you something to be completely biased?

Stating that the quality differences between the encoders is subjective is the absolute most OBJECTIVE position that one could possibly take!

Funny you should say that I am biased since I am the ONLY one here who is not siding with either encoder. The only thing I ever said was that neither encoder is clearly superior to the other. I even noted two instances where I think TMPGenc is superior in quality to cce (DV and mpeg1.) I don't know how I can be more objective than that. YOU are the one who is claiming that TMPGenc is superior to CCE in all situations even though you apparantly have never even used CCE. If that's not biased then I don't know what is.

I too am out of here though. Its no fun arguing with someone who just makes stuff up as he goes along.

[Martyn1980]

Usually when i try to compare Tmpeg to CCE, i encode first using Tmpeg, take the average bitrate, then use that as my average bitrate for CCE, and i end up with two files of the same filesize and i can then subjectively compare the two files, the end filesize is what matters, if you want the file on some sort of media.

As for a trailer being impossible, why is it, im not talking like i want a trailer to fit a CD, even CCE at 15mb a min probably wouldn't be enough, when i encode a trailer i usually stick a 100min AVI in Fit CD, set it for 2 80min CDs and use the average bitrate for that, it could be any bitrate but probably under 2520 to keep it in SVCD standards.

As for the sample looking better on a HDTV havn't got he priviledge but i usuallu crank the resolution up to 1280x1024 and jam my nose against the monitor checking for even the slightest defect, and there are more artifacrs in the CCE clip, but are they noticable on an average TV, im talking all round picture, the color looks washed out on the Tmpeg clip and its far too bright, if i turn the brightness up on my TV, then everything looks a little bit smoother.

Now i cannot believe you expect me and everyone else to believe you think Proof of Life looks exactly the same on 2 CD's as it does on the original DVD, so thousands of dollars of equipment and up to 17gb or space is wasted when you can get the same quality at 1.6gb, i have Proof Of Life ripped on this computer and its 4.71gb, damn Warner Brothers wasted 3gb, someone tell them about Tmpeg and its CQ mode, how about telling Columbia Tristar how they dont need 2 DVD's for Lord Of The Rings, they can get the movie and all the extras on one Dual Layer DVD, without any quality loss, DVD's are compressed enough, you cant keep compressing it and not loose quality, and Proof Of Life only has a max bitrate of 7.30mb, but thats the Pal version dont know if the NTSC version differs, i admited that the Proof Of Life clip looked great but not DVD quality, near considering the Filesize, and i doubt it we used the exact same bitrates as the original DVD and used any encoder then it still wouldn't match the original; which is encoded with very expensive hardware encoders.

And im in that 10% that can play Mpeg1 very well, well not at all over 352X288, but can play anything in Mpeg2.

Damn i type too much at times.

[adam]

Actually a trailer is the perfect source for a test encode. Just long enough to provide a decent amount material to analyze yet not too long for someone to have to download. There is always a good mix of low motion/detail scenes and high motion/detail scenes and the quick scene changes is a good test for the encoder's motion search ability. You can also see how the encoder handles scenes that require only minimal amounts of bitrate like the end logo/credits.

[kwag]

Martyn1980, I think you're not reading the articles in a very good detail.
Yes, I know the TMPEG sample was brighter. Here's what I said in the post:

"Now if you look at both mpeg files with WinDVD, you'll see that CCE's mpeg is darker, an advantage over TMPEG for pointing out "mosquito effects" and artifacts. I believe the luminance level was set at 0-255 instead of 16-235 when it was encoded. Again, an advantage for CCE, because artifacts are damped with this color space. "

Didn't you read that?

About the wasted space on DVD. Yes. In some cases it is. It all boils down to "Money". The industry doesn't care about puting a movie in a .15 cent CD, if they can sell you the same movie for $20.00 in a DVD. Of course you have the extras and all that in the DVD. But many people, as myself, don't care about the extras. I'm very happy with the movie an chapters.

And as for "Proof of Life", yes, it looks just like the DVD with the new 704x480 and viewed on a HDTV. And I'm not the only one saying it

-kwag

[kwag]

Actually a trailer is the perfect source for a test encode. Just long enough to provide a decent amount material to analyze yet not too long for someone to have to download. There is always a good mix of low motion/detail scenes and high motion/detail scenes and the quick scene changes is a good test for the encoder's motion search ability. You can also see how the encoder handles scenes that require only minimal amounts of bitrate like the end logo/credits.

If you have for example a 5 minute trailer, you CAN'T calculate an average bit rate for a target without a reference point! For what target size are you going to calculate? 5MB?, 10MB?, 15MB? Of course the smallest target will give you the best quality. This whole thread is about encoding modes. So if you want to encode a short trailer with CCE's 3 pass mode, what's going to be your targer size?. Because we're comparing CQ vs. x-pass, you have to use FIRST the CQ mode encoder to see what the file size of the trailer will be. THEN depending on that size, encode with your CCE using the average bit rate to try and produce the same file size as the one encoded with the CQ encoder. In this case, TMPEG. Then you can say that the test files are ready to be analyzed.

-kwag

[Martyn1980]

Kwag i did read your post i never said you didn't say it was lighter, i was just saying, You said CCE was darker, CCE produces the same as the source Tmpeg makes it lighter than the original thats what i dont like, i need to change the settings on my TV so that the borders look a proper black and not a dark grey, i dont have that problem with CCE.

As for the trailer thing, say i encode a trailer, it could be anything, but its has to be a vob file, no avi, quicktime, and its likely you would get it off the DVD of the same name, i would encode that trailer with the same Min Average and Maximum settings as the actual film,as you said use the CQ first to estatablish an average bitrate, then use that in CCE ,and theres always testing it against, Tmpeg Plus new 2-Pass which i have and i purchased so im not a Tmpeg hater, i wish i had some webspace cause then i could show my results, but im sorting that right now.

DVD's arent overusing DVD space i can see that with my own eyes i could see Proof Of Life was lower bitrate before i checked it, and Mission Impossible 2 which maintains a very high bitrate throughout looks amazing, and i didn't need no bitrate meter to know which was higher, was the clip you uploaded from this version of Proof Of Life you say is indistinguishable from the original, because im comparing the clip from the same bit on the original Pal DVD and it nowhere near, the color is wrong, there are slight jagged edges that arent on the original and its lacking some detail, and this is from a very Low Motion Scene, i will be using your template a lot, as it looks a great way for me to get great quality on my Toshiba SD210E which plays Mpeg2 out of sync, but on the Yukai i will be sticking to CCE 4-Pass Min300 Max4000.



Going to run a few tests now using Bandits trailer, which i think is nice quality, using Min 300 Maximum 4000 and an average of 1554, this will get Bandits on 2 CD's, if Amazon are correct on the running time of 123mins for the Pal Version, i will also test it against Tmpegs new 2-Pass and test it against CQ average bitrate, i will get back to you all if anyones still interested.

[kwag]

Ahhh, hell Martyn1980, If your DVD is PAL then of course you'll see a difference! Your DVD is 720x576 against my NTSC 720x480. PAL is way better, so no contest there

-kwag

[vhelp]

Afternoon martyn1980,

In reference to your "lighter" vs. "darker" issue w/ cce vs. tmpg,
Ok, first, by default, cce uses the 0-255 color space, hence the
"darker" look.
Ok, but in tmpg, by default, it uses the 16-235 color space. In
order to yield the same results of "darker" look that cce uses in
it's encodes, you have to go to tmpg's:
Setting/QuantizeMatrix/[x] Output YUV data... and be sure
it's turn on with the "x" or checkmark. I've ben this route. I've
compared the quality of cce vs. tmpg's "lighter" vs. "darker" issue
and I could not tell them appart.
Oh, and before you and anyone else here say otherwise, do this
then, in cce, select 16-235 color and re-encode. Yes, it will look
just like tmpg's unchecked [x] Output YUV data... setting.
I know, cause I've run dozens and well, hundreds of tests w/ the
above correct use and both video's come out the same color space!
In fact, the NEW sample clip I UPloaded today, I did in cce w/
both 16-235 and 0-255 color space settings vs. tmpg's
[x] Output YUV data... and [ ] Output YUV data...
unchecked settings, and when both played, they both came out the
same! To see my latest UPload sample clip, just visit my
"VHELP's Samples..." page (or not)
I've set up a template in CCE for "cvd.1850x600x2340.16-235" and
"cvd.1850x600x2340.0-255" (and a few other settings)
So, basically, this "lighter" vs. "darker" issue w/ cce vs. tmpg
is mute!

adam/kwag/marty, others, I'm w/you all on the trailer project issue.
Kwag, I think that the:
>> target without a reference point! For what target size are you
>> going to calculate?
issue is simple. Just use/assume you want to use those 80mb CD-Rs
as your "target" or "reference", ...and the game is on, he, he...
It's that simple, instead of a 650mb or 700mb CDr that you've ben
using as your "target"-to-"reference" point, you use a 80mb
CDr as your "target"-to-"reference" point. . . or, am I missing
something here, cause it just doesn't make any sense to otherwise
So, those 5min Trailor's would make perfect "target" or "reference"
numbers for a 80mb CDr, don't you all think or agree??

** note, made a cople of mistakes on those CDr's. I ment those NEW
3" or so CDr's that hold approx 80MB or so of data.

Also, to narrow down a GOOD trailer sample, we should all throw
out a few POPULAR DVD movies that have meaningful trailers to
test with, ie, here is a short list to get you rollen:
* After Shock
* AI
* Blow out
* Blue Streak
* Cast away
* Chasing AMY
* Fifth Element
* Jaws II
* Jay and silent bob strike back
* Made in america
* Millennium - has a good 2 minute trailer of it'self.
* Miss congeniality
* Ready to rumble
* Sharky's machine
* Solder
* Spaceballs
* Star wars "phantom menace"
* Suicide kings
* The crush
* The magnificent seven
* The Mummy
* The Mummy Returns
* THE ONE!
* Zoolander
* Much, much, mucho more!!

Any takes on above movies?

-vhelp
--------------------------------
VHELP's Samples. . .

[Martyn1980]

Yes I know Pal is better Kwag, but it isnt as much as you state.

Thank you for the information Vhelp, at least now i can run fairer tests and i have done anyway, i used the Pal Bandits trailer, and encoded it first using Kwags new CQ Pal Template, then i worked out an averge bitrate for the whole film, and used that for the trailer, and encoded it again using CCE 3 pass VBR on 300min 1543 average 4000 maximum, and then on Tmpegs New 2 Pass using the same bitrates, i then took the average bitrate off the Kwag CQ version which was 1722, and then encode on CCE 4 pass again using the new average, i did this so i could test Tmpegs VBR against, CCEs and also test CCEs against Tmpegs CQ, i did all these tests without watching any of the clips,the files sizes were
Tmpeg 2 pass - 28.2mb
CCE 4 Pass - 28.8mb
Tmpeg CQ - 31.2mb
CCE 4 Pass using Tmpegs CQ Average Bitrate - 31.6mb

First i watched the original VOb then the encodes

The first Tmpeg using 2 pass VBR looked pretty dodgy, too blocky, and not very clear.

The CCE 4 pass didn't look much better, less blocks but still not very clear

The Tmpeg CQ was much better, obtaining the original sharpness of the vob file, but far too blocky, every time it changed scene it would go blocky, and fast motion had the same effect, but on Low Motion stuff, fantastic, sharpest i think ive seen.

I then watched the CCE 4 pass using the same average bitrate as the Tmpeg CQ version, and i instantly could see it didn't have the same sharpness, but there wasn't a block in sight, not even in high motion stuff

So ive actually got nowhere and have came to the conclusion that, people shouldn't be so close minded and just use what gives the best results for whatever they're doing and whatever they're personal tastes are.

What i would like to know is from the Tmpeg experts is what did i do wrong, would increasing the Motion Search Precision to Highest reduce those blocks or have i did something else wrong, i do not like the blocks, and to the CCE experts do any of you know how to get that kind of sharpness as Kwags New template gives, maybe im looking for the best of both worlds, and i have to choose which i prefer and what type of film im encoding.

Vhelp nice sample, nice quality, nice girl, better than Russel Crowe and Keanu Reeves anyway, also dont have any of they trailers but i imagine i could find them.

[Martyn1980]

Trailer wise What i have on here right now is

Unbreakable
Urban Legends 2
Evolution
Forsaken
Ghosts Of Mars
Maximum Risk
Jurassic Park
The Lost World
Jurassic Park 3
Rush Hour 2
Thunderball
The World Is Not Enough
Replicant
The Glass House
The Order
Hannibal
Crouching Tiger Hidden Dragon
The Sixth Sense
Jeepers Creepers
Alien
Gladiator

[Martyn1980]

Me again guys, just rewatched in all the bandits samples in Win DVD and Power DVD, and came to some conclusions, Tmpeg 2 pass VBR old or new, is crap, too slow and crap quality, blocky and rough looking, CCE 4 Pass, looked better at same bitrates, and 1 pass does too(Thought I'd Be fair) Tmpeg using Kwags new template looks really smooth, in Low Motion, but breaks up into big blocks as soon as anything moves, the high action bit at the end looks terrible, but the talking bits look fantastic, CCE 4 pass at same filesize has a slight, fuzzy look about it, the Anti Noise does make a difference, i had it turned off, but there were no big blocks and only a few little ones at the end, plus there was a lot less noise round the writing on the CCE version, i give up, i like both.

[vhelp]

martyn1980,

The sample LIST of DVDs was just from those few that were laying right
in front of me. Now, I can't remember EVERY trailer I've seen in just
those (via the list) but maybe a few like "Millennium" cause I
specifically remember trying out Kwag's template on it, way back
when - oh the fun of trying out new things... anyway.

Thanks for the comment on my sample clip.

>> The Tmpeg CQ was much better, obtaining the original sharpness of >> the vob file, but far too blocky, every time it changed scene it
>> would go blocky, and fast motion had the same effect, but on Low
>> Motion stuff, fantastic, sharpest i think ive seen.
Ok, here is where I can probably answer (in short) and to the best
of my understanding of this same finominom (spelling)
* The reason why you see all those blocks in TMPG's CQ encodes aka,
Kwags template, during those fast scenes/changes is bacause of those
the long GOP settings. Yes, this distortion is when the scene
changes and while tmpg's cq is alorigically processing the frames,
(in short) the distortion is the result. I'm sorry i can't explain
it in detail here, but I know for a fact the above is the reason,
but that I can't explain it correctly enough for you to understand!
It's all in my head. Really!! It is! But, maybe Kwag can MUCH
better explain it (in all honesty) why this happens in some encodes.
But, this is the reason (above)

** wow, I just realised that I UPloaded the WRONG clip! he, he...
on this clip, you can see a slight "white"ness in TRIO's logo at
bottom right. Final filesize was suppose to be 12.6MB, oh, well.

oh, and keep the posts comming. . .

-vhelp

[kwag]

Me again guys, just rewatched in all the bandits samples in Win DVD and Power DVD, and came to some conclusions, Tmpeg 2 pass VBR old or new, is crap, too slow and crap quality, blocky and rough looking, CCE 4 Pass, looked better at same bitrates, and 1 pass does too(Thought I'd Be fair) Tmpeg using Kwags new template looks really smooth, in Low Motion, but breaks up into big blocks as soon as anything moves, the high action bit at the end looks terrible, but the talking bits look fantastic, CCE 4 pass at same filesize has a slight, fuzzy look about it, the Anti Noise does make a difference, i had it turned off, but there were no big blocks and only a few little ones at the end, plus there was a lot less noise round the writing on the CCE version, i give up, i like both.

Of course they breaked up You encoded with the KVCD template that has a MAX bit rate of 2,300 and you encoded your CCE sample at MAX=4000 !. Do it again at a MAX of 2,300. Then compare your CCE again and look at your blocks again. They'll be worse with CCE.

-kwag

[vhelp]

marty,

I have to agree on this issue w/ kwag cause by accident, on one of my
encoding projects w/ cce vs. tmpg (yesterday) I did an enocode w/
framevering to *.avs --> vdub --> cce.
Ok, after viewing my cce encode, I nothice that it was sharper and more
filled, but I couldn't get over the blocks in some fast scenes. I was sure
that it was my eyes, till I realised that I DID NOT LOAD my *.mcf template
of filters in vdub. So, what happend was cce encoded my clip at 720x480.
I have a resize param 720x480 down to 352x480.
* i was experimenting w/ 720x480 captures for cce

Anyways, only reason why I mentioned above w/ detail is because I just
so happend to be using 2340 as my bitrate, and at 720x480, the blocks
showed up on fast scenes pretty well, using 4 total passes! But, yes,
raising the max to 4000 should eliminate it, but then, you would have a
larger file, and if you're trying to squeez a much MBs on a CD, then you
really have to comprimise a little, you know? But, I suppose that 4000
is ok for small clips vs. whole movies though. so, you may be OK on this
one though. Just my 2 cents worth on the issue of CQ/VBR, etc vs.
fitting as much movie per CD vs. quality, etc., etc.

On another note, I've ben thinking about that trailer issue. Actually, why
do we even need to use a trailer?? Why not just pick a good 5min. scene
w/ some slow to fast movements. Don't really need a trailer
for this. If time is crushial, why not make it a 10min. instead or something
like that.
Well, I thought it worth mentioning.

-vhelp

[kwag]

Because this thread is a discussion about CQ vs. X-pass VBR, I've done "correctly" what everyone wants to see. With samples provided so that you can try and match them. 8)

The trailer encoded is from the movie, yet to be released, xXx. No, it's not a porn movie 8) but that's the title xXx.

Here is the original file:
http://ns1.shidima.com/kwag/xXx_DivX.avi

It's about 10MB in size, excelent quality, and playing time of 90 seconds.
The resolution of the avi is 512x256 and a frame rate of 23.976.
I have encoded this trailer with both CCE ( I borrowed a copy now for the test, because my friend's office is closed. So I can legally use it on one machine at a time ) , and also with TMPEG.

Both samples were encoded with the same AviSynth script here:

LoadPlugin("C:\encoding\MPEG2DEC.dll")
AviSource("c:\download\xXx_DivX.avi")
BicubicResize(704,352,0,0.6,0,0,512,256)
#TemporalSmoother(2,2)
AddBorders(0,64,0,64)

As you can see, I have used Bicubic resize, to upscale the resolution of the original AVI from 512x256 to 704x480 (with borders). This makes matters worse for both encoders, in order to bring out the little details of noise etc. and see how both encoders cope with this source 8)
The AviSynth script was processed directly with TMPEG and "KVCD 704x480 Plus" template. The only changes to the template were the following, to match the parameters of CCE.
Changed encoding mode to MPEG-2, Because the template's default is MPEG-1.
Changed the VBV to 112(which is 224KB) to match CCE's value.
Changed precision level to 9 bits. Just like in CCE.

Now the test. I first encoded the KVCD version, with it's default CQ_VBR and GOP, processing the .avs script directly. The file size is 21,026KB.
Next I took the average bit rate from that sample with bit rate viewer, and encoded the AVI with CCE. The file size was WAY lower and looked like crap. I don't even remember what it was. So I had to start increasing the average value until my encode was as close as possible to the size that the CQ mode created with TMPEG. The value for average bit rate in CCE is 1,950Kbps. The file size for the CCE encoded file is 21,145KB. Close enough, with a slight advantage over TMPEG, that can be considered null. 100KB in 20MB is nothing.

So the values for CCE are MIN 300, AVG 1,950 and MAX 2,300.
For TMPEG, it's the same MIN 300, MAX 2,300 and there's no average known in advanced, because it encodes in CQ mode. So we know the average when the encode is done. The encoding was done in CQ_VBR mode with a value of 15, using KVCD's GOP and Q. Matrix.

For CCE's encoding, I saved the project from TMPEG as a .tpr, and created a pseudo .AVI with VFAPI, to be able to read the file with CCE. Because you can't read an .avs directly with CCE, and I needed to process the DivX file via AviSynth in order to create the correct resizing, etc. Also note that both samples have an extension of .mpg, so that you can open them with WinDVD, but they are actually video streams only. There's no sound. Only the original DivX has sound.

So here are the results, for your evaluation.

CCE's sample:
http://ns1.shidima.com/kwag/xxx-CCE-3pass.mpg
TMPGEnc's sample:
http://ns1.shidima.com/kwag/xxx-TMPEG-CQ-VBR.mpg

I think there's no contest as to what mode is better 8)

If anyone can come up with a better quality than TMPEG's CQ mode, IN THE SAME SIZE as the sample, with any encoder in 3, 4, 5, 20 passes etc, I'll be happy to dump TMPEG for that encoder.
Fair enough? Remember, to be fair, you HAVE to produce a file size as close as the one encoded with CQ mode. I even SEE a better motion estimation algorithm with TMPEG over CCE, the difference is very VERY small, but it's there. But the difference in price is HUGE. As I said before. CCE has lost it's edge. Is X-pass better than CQ. NO WAY. The proof is here, and that's what all this thread has been about, and hopefully it will end soon. If anyone is still not convinced, then let's do it on another trailer!. The results will be the same over and over again.

-kwag

[adam]

If you have for example a 5 minute trailer, you CAN'T calculate an average bit rate for a target without a reference point! For what target size are you going to calculate? 5MB?, 10MB?, 15MB? Of course the smallest target will give you the best quality. This whole thread is about encoding modes. So if you want to encode a short trailer with CCE's 3 pass mode, what's going to be your targer size?

I don't understand where your going with this. There is no target filesize. The only thing that matters is that both tests use the same bitrate settings. The most logical thing to do is to make a number of tests at different bitrate settings to test the quality of each encoder at various bitrates. For instance TMPGenc might seem to provide higher quality under 1mbit avg while cce might perform better above this level.

Its not difficult to come up with typical bitrate settings even when your file is very short. For instance make several tests using typical CQ % levels such as say, 60, 65, or 70. Go ahead and throw in a very high level like 85 and a very low one like 30. Then go back and calculate the avg in each encode and use it in the multipass encodes.

Because we're comparing CQ vs. x-pass, you have to use FIRST the CQ mode encoder to see what the file size of the trailer will be. THEN depending on that size, encode with your CCE using the average bit rate to try and produce the same file size as the one encoded with the CQ encoder. In this case, TMPEG. Then you can say that the test files are ready to be analyzed.

Yes of course, doesnt this go without saying? Obviously this is the only way to ensure that both test encodes use the same bitrate settings. This is what I have been saying since page 1.

In regards to some of the other posts...I do not think that any test should use non-standard settings such as a resolution of 704x480 or a max of 4000kbits. Besides the fact that not everyone will be able to test these on their hardware players (where ANY subjective test viewing should be done) these do not represent typical encoding settings. Sure alot of people use non-standard settings, but I think the majority of people prefer to at least attempt to make their svcds compliant.

As far as the modified gop settings...its up to you. If you use them in TMGenc then you are comparing TMPGEnc w/ modified gop vs cce, not just the encoders themselves.

All tests should use mpeg2. I think it will be difficult to find anyone who prefers cce over TMPGenc for mpeg1, including the manufacturers of cce oddly enough.

As far as the luminance settings I think everyone should read this post, http://forum.doom9.org/showthread.php?s=&threadid=28815 and ensure that both encoders are setup equally.

I do not think ANY noise reduction should be used, either in the frameserv script or in the encoder. The point of an encoder is to preserve as much of the source's quality as possible at a given level of compression. While some noise reduction may be useful in real world settings, for the purpose of comparing the encoders it is best to give it the highest quality, most unmodified source as possible. The point of this test, as far as I can tell, is to test the quality of output of cce vs TMPGenc, not the quality of their noise reduction filters or how any avisyth noise reduction affects the quality of cce's or TMPGenc's output. The only thing the frameserv script should contain is the source input line and the cropping and resizing commands, and any non-quality commands such as a resample audio line for cce.

Ive said it before but I'll say it again, I do not think its possible to make a truly fair objective test between the two encoders. I think the importance of the "image quality priority" setting is being vastly overlooked. You cannot simply use the default setting, that is setup to be used for the default bitrate setting which is 6000kbits! The lower avg bitrate you use the higher priority setting (slider further right) you have to use otherwise you get mosquito noise, which seems to be a recurring complaint of cce's quality in this thread. And if your priority is set too low (slider too far left) at any bitrate than mosquito noise will appear. This is one variable that cannot be eliminated because there is no way to determine a setting which is comparable to the one that TMPGenc would use in its interal algorithms. This is why all tests between the two encoders amount to nothing other than a subjective comparison, which is epitomized by the fact that given the same samples Martyn1980 preferred one while Kwag preferred the other.

Of course all of these suggestions are just my opinion and only apply given the assumption that this test is meant to determine the highest quality encoder for SVCDS. It also might be viable to test the same source at much higher bitrates as well, such as avg 6mbits. It is at this point that I think cce's strength as an encoder really shines.

[kwag]

As I said in my previous post Adam. There's the original DivX file. Try to match it with CCE at any settings with the same constraints ( resolution, frame rate, file size). Those are constant values, so that you can play with all the intrinsic variables of the encoder that you want, and see if CCE can be optimized above TMPEG's CQ modes with X-pass VBR. The bottom line, and I continue to sustain what I've always said, is that CQ modes create higher quality than X-pass VBR modes. CQ is not equal to VBR. It's a completely different mathematical encoding algorithm.

-kwag

[adam]

Kwag, not that your test isnt valid, because I think it is to some extent, but I do think it can be better.

Divx is a horrible source to use in a test encode. It is inherantly different from mpeg2 and doesn't transcode well to any other format. The fact that your source has already been transcoded at least twice (once from film to vob or mov and then again to divx...No production studio is going to release a movie trailer directly in divx.) means that it is going to have a number of artifacts. Since the purpose of an encoder is to preserve quality the better encoder will actually preserve MORE of the artifacts. Since the resolution was actually increased (never preferable under any circumstance) these artifacts will be even more apparant, further swaying the results.

Also I just posted the above post before even reading yours but I think alot of those points are valid. I would be more than happy to do a challenge of sorts, and attempt to author an mpeg in cce that looks better than your TMPGenc file but now is not a good time since I am in the middle of moving to another city. However if this were to be done we would have to use the exact same source and it would have to be of high quality. The only way I can see that this would be possible would be if we used the same trailer taken off the same dvd.

[kwag]

Well Adam, I've already tried VOB's and satellite captures as source, and the results are the same. That's why I had the courage to post what I did. I wouldn't have posted any tests without doing trials myself, and double checking that what I was about to post, wouldn't be nonsense! I am aware that DivX is a couple of generations of encoding, and when we re-convert it again to mpeg(1,2,xyz), the result is worse than the original. But the result is as valid with a cleaner source. Because if you have X amount of noise figure in a material ( analog noise, macroblock, etc ), both encoders will see the same thing, and encode what's thrown at them. For some reason CCE produces a "grainier" result in the mpeg files, even when the source is very clean. This of course is not viewable in regular TV's, and not even in HDTV's, unless you start to do "pause" on different scenes and stick your face 12" from the CRT! Also I could tell on the "Matrix" samples, I paused on some of the action scenes, and on the CCE sample you could hardly distinguish details in the face of Neo. But on the TMPEG sample, they were there. Still blurred in the action, but the details were there. So I think that the motion estimation in TMPEG is superior to the one used by CCE. I think that people haven't really done that many tests with the latest versions of TMPEG, and probably they're comparing the results of older versions of TMPEG, which were much lower in quality compared to CCE. I've seen the quality enhancements of TMPEG, and it just gets better on every version, but for CCE, I haven't seen any quality increase in the 2.x branch. I believe people do report faster encoding with the latest versions, but I haven't done any tests myself. I wish TMPEG was faster, but I wont trade quality for encoding speed. As far as multi passes, I really think that the only advantage is that the target size can be predicted very closely, and with CQ you can't. But if i want to encode a two hour movie right now, with the KVCD 704x480, I don't have to go through X-passes because I know that 99% of the times most 120+- minute films will now fit in two CD's.
And if it doesn't, then I can either boost the CQ up and re-encode one more time to fit it in three CD's(which I wont), or lower the CQ just a tad so that on the second encode it will fit. And that's what I call "Manual second pass CQ VBR"

-kwag

[adam]

Ok Kwag if to your eyes the TMPGenc clips always look better than you have obviously proven to YOURSELF that TMPGenc is better. I just think you need to realize the subjectiveness of it all, yes I did say that word again. You just need to realize that you cannot claim that TMPGenc is undeniably and inherantly better quality than cce, anymore than you can say the opposite is true. Also I have yet to actually see an even remotely fair comparison between the two done by you or anyone else. Ok you say you have conducted one but no one has seen it yet.

Giving a noisy source to both encoders and seeing which one results in the better qualty does not show which encoder is higher quality, it shows which encoder handles noisy sources better, which may be a valid bit of information but neverthess that is not the point of this thread. But even this is an oxymoron because less noise means MORE of the original quality was removed.

If you want to do a fair encode you need the highest quality source possible and you need to remove as many variables as possible. I really think my previous suggestions are valid in this regard.

I must say I find alot of your statements quite silly. I can't see how you can claim that extra passes is only used to predict filesizes. Forget about every document ever written about mpeg and the fact that the ability to encode more than 2 passes even exists (If filesize can be predicted exactly in 2 passes why scan it again? Or again another 2 times?) why not just test it out for yourself? Anyone attempting to pit one encoder against another is biting off more than they can chew, its impossible to prove anything as fact. Why not test something that CAN be proven as fact without a shadow of a doubt? In either TMPGenc or CCE do one encode at 2 passes and one at 3 and analyze the results. The 3 pass vbr encoded file will ALWAYS have a lower Q value and have more intelligent bitrate allocation, assuming your bits per pixel is less than the source otherwise VBR is pointless anyway. I have no problem admitting that TMPGenc can outperform cce in certain circumstances, I have already admitted that. What I do have a problem with is when proven facts are simply disregarded. I think a much more valuable test, for you at least, would be to verify what every document on mpeg encoding states, that more passes does have an affect on quality.

For the record I never suggested that CQ vs x-pass vbr be tested by pitting cce against TMPGenc. I personally think this makes no sense. If one encoding mode seems to outperform the other how do you know its not just the encoder itself? You very well may be able to convince people that CQ outperforms CCE's multipass encoding in quality, though my personal tests do not show this to my eyes. But if you simply use common sense and use those facts which can actually be tested and proven, than x-pass vbr will ALWAYS outperform 1-pass vbr. And regardless of what you believe CQ is simply 1-pass vbr.

I'm not interested in proving anything as fact, I simply think its only fair to the readers of this forum that you and anyone else only make reasonable claims. If you cut the bitrate of the source in half than there is no way it will be just as good no matter what template you use. I think you need to be a little more honest about the results your template provides and also realize that its only a template. The encoder is the real one doing the work. You should also not claim that you have proven one encoder superior to another, since this cannot be proven objectively. The FACT is that TMPGenc and cce are both very good mpeg2 encoders and that either one should be able to provide more than acceptable results. Depending on the source and the settings used either encoder has the potential to outperform the other. I really think it should be left at that.

[vitualis]

@ kwag:

I strongly suggest that you read adam's original criticism to your "test".

If you think that you have proved anything then you are very misguided.

If you are trying to compare which encoder has better quality, then all other variables must be constant.

For instance, they must first be encoded to the SAME thing! That is, both MPEG-1 or both MPEG-2.

The overall bitrate settings must be identical (min, max, average) -- that is, the same file size.

You MUST look at the encoding in its ENTIRETY. Looking at a small clip is absolutely meaningless. It is like saying (for example) one necklace looks better than another by examining only one link. The average Q for any one section for the movie may be higher for one encoder than another. That, however, is meaningless. It is the AVERAGE Q for the entire clip that is important.

I'm sorry, but you haven't proved anything and trying to insist that you have only looks silly. Furthermore, several people on this thread have already disagreed with your interpretation and subjective impression of quality of those clips.

If you want to do an objective test (a worthy objective), then do what adam has recommended. Use the SAME settings for both encoders. Encode the SAME material. And then, analyse the ENTIRE clip.

Regards.

[kwag]

The only silly people here are the ones who talk talk talk without any concrete samples for evaluation. I posted a trailer. And I posted the samples. The MPEG-2 CQ mode made with TMPEG is better than the CCE. Dont give me the crap that it's "subjective" or "objective". It's just looks better and has less artifacts than the CCE encode. It's really funny that it's always the same two or three guys who just keep on hammering the same ideas, but NEVER provide any proof.
Now I'll say it again. PROVE that CCE is a better encoder. I encoded from the same source with TMPEG and CCE as MPEG-2, and the TMPEG sample looks better than CCE's X-pass MPEG-2. Now what part of that don't you understand, ah!
Or are you too in CCE's payroll?
Let's make it simpler for you, and everyone to understand. Encode the AVI sample with CCE with the best parameters and variables you have. But the final file can't be larger than the sample I provided as MPEG-2 made with TMPEG. Is that too hard to understand? Then compare the quality.

-kwag

[vitualis]

Yeah, it is funny how the same 2 or 3 guys keep posting nonsense isn't it?

It has been explained clearly why your test is flawed. There is no point looking at the results. It's like comparing brand A and brand B orange juice. It doesn't matter which one tastes better if one of them is actually apple juice. It has no bearing on whether A or B has better oranges.

And even if you did, several people on this thread have already mentioned that they like the CCE encoded clip better. Subjectiveness is already at play. Your assertion that the TMPGEnc clip is "better" is no more than a subjective opinion that not everyone agrees with... not that they can even be compared.

I'm not trying to prove CCE is better than TMPGEnc. I really don't care all that much. But YOU are trying to do the opposite -- and you can't convince me or anyone else with absolutely shitty methodology. You haven't got any evidence. Comparing your two clips is like comparing apples to oranges.

If you really care all that much, then do the test properly. Same format, same bitrate settings, same filesize, same clip.

As for CQ, of course it is nothing more than 1-pass VBR encoding. Sure, the bitrate distribution algorithm is different, but it is still 1-pass, and VBR encoding. It has been explained plenty of times before why multipass VBR encoding is always more efficient (on the same encoder at least and according to the caveats I posted back on page 4).

Why can't you people realise that CQ, multipass, etc., are ALL ONLY ENCODING THE SAME THING -- MPEG frames. If all other things are equal, what is different between these modes is only in bitrate distribution. Multipass VBR (if the encoder is not BROKEN -- read the caveats I posted before) will distribute the same number of bits (same filesize = same measuring stick = only fair way of comparison) more intellgently than ANY one-pass (CBR, 1-pass VBR, CQ) method.

Regards.

[Martyn1980]

Damn i go to bed, i go shopping and look what ive missed.

Keag just downloading your clips now and then im gonna do a test encode myself and i wish i could back up what i have but i dont have the resources to keep uploading Mpeg samples.

I myself have said CCE is grainier, but on the new High Bitrate DVD's out there was a review that Crouching Tiger, was so high detail that it looked grainy at times, Tmpeg is visibly smoother, but at the expense of what, as for Me using a 4000max, why not, my DVD supports it and i might do another encode with a 2300 max but i cant be bothered sick of that trailer now, on the same subject vhelp it doesn't matter what the max bitrate is or the min, as long as i keep the average the same, the filesize will not increase, i did K pax the other day with a 300min and a 4000max and i got the whole 2 hour film on 2 discs, with great quality, and Kwag if you encode with CQ first then use then take the average bitrate using a bitrate viewer and CCE gives you a much smaller filesize then use another bitrate viewer, when i do that i feed your CQ version into Virtual Dub,as long as its Mpeg1, and it gives me an average bitrate if i use that i get nearly exactly the same filesize, give or take a few bits.

CCE does take avs scripts directly, just not the newer versions though, although i heard the latest one does, but i dont know, i personally use 2.5

And comparing Video is totally subjective, its like comparing films or music, or clothes or women, just becsuse one person like something and thinks it better doesn't mean everyone should, you can make people like what you like, and i think more people like Tmpeg because its the underdog, its the cheap encoder that only costs money because of Mpeg2 licensing and CCE is the expensive big business product, its the same with AMD vs Intel, Windows vs Linux, people should judge on what pleases them, not what other people tell them is better, and does the price matter, well how many people pay for CCE, i payed for Tmpeg though.

Im going to look at the samples now and try a few test on the original myself, if i like the original but i wont encode crap, a while back in this post someone directed me to a divx trailer, The Matrix and it was poor quality, and i wont waste my time on it.

Hey virtualis am i one of these 2 or 3 guys posting nonsense and if i am, what parts are nonsense

[injunpana]

If you want to do a fair encode you need the highest quality source possible and you need to remove as many variables as possible. I really think my previous suggestions are valid in this regard

Now this statement is what's silly. If you take this argument, up to what point in the quality of the source shall it make it a fair test?

If you only use the highest possible quality source, maybe you can even take the crappiest encoder and it will produce an encode that may be indistinguishable from the other encodes. You know, human eyes are easily fooled. The thing is , if you don't use a bitrate viewer or any other quality evaluator, the fact that the crappy encoder did a poorer encode is hidden. That's why kwag did good in using an ordinary quality source.

Maybe you forget that the the test is to see what mode reproduces closest to the original source. Whether one handles noise better than the other one is immaterial, the bottom line is, " how close is it to the original?" So any material source will do whether it is noisy or not. The only parameters that should qualify a fair trial is same output size, same min/max/ave bitrate, and same mpeg1 or mpeg2.

There are so many excuses. First, when CQ and x-pass vbr was compared in TmpGenc and CQ came out better, the outcry was, "you can't compare them that way because CQ in CCE is crap, try the CQ in CCE!" So kwag made the tests again to satisfy those people, and when it might look like CQ is better again, Vitualis comes in and says,"you can't compare them using different encoders, you have to use the same encoders!"

Now, all what's left is you have to prove your point by comparing the two in CCE. If you lose here, they will say, "see? we told you x-pass vbr is superior." But if CQ still wins out over VBR, they will just tell, "this proves nothing, it just tells you that TmpGenc and CCE handles CQ better than VBR. You can't win kwag! Stop it now!

But if you still want to continue, get all the available encoders and try the test on them all. I doubt it will change anything though.

[injunpana]

Wher have you been all this time?

For instance, they must first be encoded to the SAME thing! That is, both MPEG-1 or both MPEG-2.

kwag has been there and done that.

The overall bitrate settings must be identical (min, max, average) -- that is, the same file size.

kwag's been there and done that

You MUST look at the encoding in its ENTIRETY. Looking at a small clip is absolutely meaningless. It is like saying (for example) one necklace looks better than another by examining only one link. The average Q for any one section for the movie may be higher for one encoder than another. That, however, is meaningless. It is the AVERAGE Q for the entire clip that is important

kwag's been there and done that

I don't know kwag, they're missing a lot of what you're saying. Maybe because they're too long? Maybe you have to make them short and sweet?

The fact is they so totally hate your statement "CQ is better than VBR" that they read your post but nothing registers at all. You can accomodate all the conditions they ask but there will be something more again and again, or they will go back again to the beginning like vitualis (see the above quotes) until in the end they will just say it's like comparing apples and oranges, that there is no way to compare the two fairly. Give it up kwag!

[therick]

hhhmmmmm....why is it, that a member with 37 posts...who has been on this board roughly 2 1/2 months feels the need to throw himself into this argument between two of the smartest guys on this board and kwag, the wily newcomer who is opening eyes and changing minds on this board...

@injunpana- you comments are uneeded and make no sense...I have been on this board longer than you...and even though I started this little X-pass vs. CQ thing, I bowed out to watch...and so should you...please

[Martyn1980]

I cant speak for everyone, but myself and i imagine Adam are open to new suggestions and Adam has openly said that he uses Tmpeg for DV, and i use it myself for my Toshiba DVD player, as it dont like mpeg2 that much, from my own tests i found Tmpeg to be better a low motion stuff, and i dont care what you say about my Maximum bitrate i still stayed within the maximum filesize, increase the maximum on CQ to 4000 and tell me the filesize.

On the Divx tests in my opinion they all look shit including the original, Divx, has a similar look to Tmpeg very smooth, but lacking some detail.
Look at that trailer, its not the best out there is it, and the encodes are so similar its not worth bickering about, and if i increased the Max to 4000 or even 3000, there would be a difference in quality, why cant people just accept each encoder for what it does and each way of encoding and use whatever works best, for their source, and DVD player they're using, if like me you have a DVD player that has trouble with SVCDs then use Kwags template on Tmpeg and if also like me you have a DVD player that supports Mpeg2 up to really high bitrates, then use 4 pass VBR on CCE with a high maximum, Kwags template do amazing things with such a small bitrate but dont get ahead of yourself, your not going to get Studio quality results, using a nearly free encoder, at a third of the original filesize, which itself is heavily compressed, this whole post is going nowhere because nobody is backing down and injunpana, people only quit when they're losing, i myself will continue to use both encoders using a variety of different ways of encoding, if you lot want to just use one encoder with one way, on every single source do it, but dont expect everyone else to follow, i only want the best quality in the smallest filesize and right now it my way on CCE, Tmpegs close but not close enough, but i will keep using both, because i need to for my DVD player situation.

[wway]

hhhmmmmm....why is it, that a member with 37 posts...who has been on this board roughly 2 1/2 months feels the need to throw himself into this argument between two of the smartest guys on this board and kwag, the wily newcomer who is opening eyes and changing minds on this board...

@injunpana- you comments are uneeded and make no sense...I have been on this board longer than you...and even though I started this little X-pass vs. CQ thing, I bowed out to watch...and so should you...please

therick,

SOMETIMES people who talk the most, know the least. The number of posts is no indication of an individuals knowledge of the subject.

I have very few posts. However, I've been READING this forum and encoding for more than a year.

wway

[therick]

...and I have been reading and encoding for over a year...who cares? You probably know more than me or I know more than you...it doesn't matter-the reason I said anything is that that guy's comments don't make any sense and just make me have to scroll down further to read this post (as this post is doing ). He adds nothing new, he just comes in on the side of kwag- as it stands now I am on a side, but do you see me weighing in with unnecessary comments about kwag or virtualis or adam? Nope.