battle of the CBR/VBR/CQ modes

[vitualis]

vitualis
I do a few test encodes before the final one. I aim to use most of the disc-space needed to suit my quality setting, but if I don't completely fill every disc I do not lose sleep over it.

Fair enough... However, if you are doing test/multiple encodes (essentially multipass), isn't it more logical to do a simple multipass VBR? (time to ditch TMPGEnc perhaps?)

My aim throughout has been to counter those who choose to rubbish CQ, pointing out that 2-pass VBR (at least with TMPGEnc) can give a poor result if you choose too low an average bitrate. This is more likely to happen if you are hung up about the number of discs you actually use. I never said multipass VBR was rubbish or was not capable of giving the highest quality results. My aim is not to evangelise - merely to counter misinformation. 8)

Again, fair enough. I've pointed out quite early on that CQ with TMPGEnc may well offer better quality than its 2-pass VBR (technical issues with TMPGEnc). However, what I've tried to argue is that multipass VBR (that is, 2 pass plus) is theorectically better than CQ encoding -- and this is true with the other encoders.

Sure, a too low an average bitrate with any pass VBR will yield poor results, but what I've tried to counter is that the same is true for CQ. An inappropriate Q co-efficient will lead to poor results too. It ultimately comes down to experience and on this issue, I don't believe this is a benefit of either encoding method.

What I've mainly tried explain is how VBR encoding in its various guises (CQ, multipass, etc) actually works. It has been strongly suggested by various people on this thread (though perhaps not by you directly on perusal of your posts) that CQ offers superior video quality to multipass VBR on theorectical grounds. That is, for the same size file, any encoder. I've explained many times previously why this is not the case.

For the same file size, same video clip, same general bitrate specs (max, min) and taking into account all the caveats I posted before, multipass VBR should offer better quality than a CQ clip.

As for disc usage, let us assume 80min discs for now (and using SVCD specs).

• for anything < 40min, maximum of 1 disc @ CBR 2500
• for anything < 80min, maximum of 2 discs @ CBR 2500
• for anything < 120min, maximum of 3 discs @ CBR 2500

For example, if your video clip is only 100 minutes long (1 hour 40 min), you should not be using VBR in any form and 3 discs. If you are using three discs, you should simply be using CBR 2500.

That is, if you video clip is 100 minutes long, you should definitely be aiming for 2 discs and then work to fill those discs. 1 disc is possible, but only if you are crazy...

If you work through a fairly simple algorithm of this nature, it should be relatively obvious how many discs to use for any length video clip.

Regards.

[vitualis]

Actually if you are trying to see which one will encode with the least amount of artifacting, then a dodgy DivX would be a good choice. Most encoders will handle clean sources well, but give them a noisy signal and they will artifact all over the place.

Actually no.

You need a clean source (DVD is good, a digitally generated clip may be better).

If a source is noisy to begin with, you are adding an unknown element and the results will be hard to interpret -- e.g., is the end result better because the encoder does some sort of prefiltering? This has nothing to do the efficiency or quality of the encoder itself and furthermore, unintellegent filtering can WORSEN the end result from a high quality source.

Regards.

[energy80s]

Michael can't see what you are talking about. Looks like you just want to use CBR for everything - which is fine as it's what I do most of the time.

As for the noisy signal problem, I stand by my assertion that a noisy signal is more difficult to encode and therefore a better test of an encoders abilities.

[injunpana]

All respect injunpana but what exactly do u call stunning and near DVD quality, maybe im too fussy but the only time ive ever thought of an encode as near DVD quality was when i was bored and encoded a couple of trailers at 5000kbps CBR, if i see one block, any grainyness, blurring, then im not happy

Near dvd quality meaning it's devoid of blocks, no perceptible graininess or blurring even with your face against the monitor. As I have said, I'm using the kwag 704x480-1cd template modified a little bit. Bitrate settings were 300-2300kbps CQ=74 and audio is 48khz 192kbps. The avysinth temporal smoother was set to 1,2 but you can do away with using it and you wouldn't know the difference. It came up with a file size of 1,457,453kb. I could still raise the CQ to optimize the 2-cdr capacity but I see no need since the quality is perfect as it is. Oh yeah, the movie is "The Matrix".

[injunpana]

Martyn1980, if you really want to see what I'm saying, why don't you do an experiment just for the heck of it.

Get the "The Matrix" or any other dvd and rip it. On TmpGenc, use the size 704x480, use the default quantize matrix, and the Kwag GOP arrangement (I believe its 1-18-3-1-48 ). Set the CQ to 74-80 depending on the length of the movie with bitrates of 300-2300kbps. You can downsample the audio or just let it be at 48khz, choose 192 or 224kbps - its your call. These settings should give you a file size that would fit 2 cdrs providing the movie is in a widescreen format. Then encode and tell me if the quality just ain't grand.

[banjazzer]

Fair enough... However, if you are doing test/multiple encodes (essentially multipass), isn't it more logical to do a simple multipass VBR? (time to ditch TMPGEnc perhaps?)

I only do a couple of test encodes on what I hope will be representative samples of the movie. If I get it wrong, my calculations will be out. However, several test encodes followed by a CQ encode is still a lot quicker than 2-pass VBR.

I have no doubt that most of those here who use multi-pass VBR know what they are doing, so I would think it unlikely they would choose an insufficient number of discs. I would expect they always get good results. I always get good results, using CQ. My personal reason for preferring CQ is because I like the saving in time. Although I think it is possible to get the filesize wrong with multi-pass VBR, that is not my main reason for preferring CQ.

I actually do have access to CCE, although I have not used it to encode. I'm sure there are ways around it, but it seemed to me that only accepting avi input is a bit of a limitation. My intermediate stage from DVD involves using an avs file, which seems icompatible with CCE.

Having said that, I have never found anything to criticise about TMPGEnc judging by the results it gives me. Maybe something a little quicker would be nice, but I would be surprised if I would notice much improvement in quality using other encoders given equivalent file sizes.

[Kdiddy]

My intermediate stage from DVD involves using an avs file, which seems icompatible with CCE.

Use version 2.5. You should not have amy problems then. Don't forget to add the ResampleAudio line if you are using an Athlon chip (regardless whether or not you are encoding audio with CCE).

[Daagar]

I agree with injunpana. Using the 704x480 template, I've gotten better quality than any other method I've tried - especially for the filesize! I'm sure if I had a 60" projection plasma HDTV 16:9 anamorphic widescreen whatever TV, sure maybe I'd see a difference. But then again, if I had all that, I'm not sure what I'd be doing here encoding stuff. I could just go buy my local Hollywood Video and be done with it.

To put some hard numbers... a recent 704x480 VCD I did was reported in bitrate viewer to have a peak of 2107kbps, an average of 788kbps, and an average Q level of 4.76. And I'm really hard pressed to find that much difference from the original source... the fact that I keep nearly full DVD resolution makes quite a huge difference. (I used a CQ of 60, 300min, 2300max, 1-18-3-1-48 GOP and the default TMPGEnc matrix).

[Martyn1980]

Just a quick thing before i start, can u use Mpeg2 with Kwags templates, cause my DVD player dont play high res Mpeg1.

I recently put The Order on 2 discs and its only a 85min film and i used 4 pass VBR, and 160kbps second audio, and i still wasn't too happy with the results.

It was great if i hadn't saw the original, but you cant tell me you cant see much difference between them, because the matrix is over 2 hours long and thers no way your going to get over an hour over an hour of DVD quality footage on a disc, what your stating is an opinion, i put bandits on 2 discs and and wasn't that happy but my girlfriend thought they looked the same as the original.

Im not disagreeing that you get great quality for the filesize, but that doesn't mean its great quality.
I used to be a 704x576 or 480, but at low bitrates it isn't really practical, in a low motion clip it looks great, nearly DVD, but high motion and it goes all messed up, ive tryed CVD which has less blocks, but im not too sure on it, and in spite of what people say you can tell the difference on a TV, and im only using a 21inch 4:3 thing.

I might try that test injunpana, i have the matrix but its hundred of miles away in my parents house, ive looked through my stuff i have here and i have Gladiator at 149mins, Hannibal at 126mins, both Region 2 Pal discs, and Thunderball NTSC which is 130mins, my video store doesn't have the matrix, and i dont want my family to post it too me, will any of these do.

[injunpana]

It was great if i hadn't saw the original, but you cant tell me you cant see much difference between them, because the matrix is over 2 hours long and thers no way your going to get over an hour over an hour of DVD quality footage on a disc, what your stating is an opinion, i put bandits on 2 discs and and wasn't that happy but my girlfriend thought they looked the same as the original.

I don't want you to take my word for it when I said I get near-dvd quality that's why I suggested you try it for yourself. You could use The Matrix or any other movie, it shouldn't matter and should work on all movies.

Of course there is a difference between the dvd and the 704x480 encode but it's so subtle that unless you have super-eyes you can't pinpoint it. If you have them playing side by side, then maybe you could see the superiority of the dvd over the compressed copy.

[injunpana]

You could try Gladiator, and being a little long, maybe you could use a CQ of 64 or 66. What I usually do is after a few minutes into the encode, say 6 minutes, I peek into the explorer and look at the file size and divide this by 6 to get an approximate mb per minute. I then multiply this by the total length of the movie to get the estimated total megabytes of the encode. I then abort and adjust the CQ to get the filesize to be around 1,600,000kb when done. The rule of thumb usually is 100mb for every drop or raise of 4 in the CQ setting.

I also have a copy of Gladiator in hand so maybe I will try to encode it tonight, then we could compare results.

[Daagar]

You can of course switch to mpeg2 for the kwag templates, though I'm not sure what it exactly does to the result. It should be equal to doing it in mpeg1, though. I personally prefer VCD, because my player requires the 'header trick' to play SVCDs. One nice thing with switching kwag's templates to mpeg2 is that you can turn on 3:2 pulldown.

Injunpana... try switching the Kwag template from the Andreas matrix back to the default TMPGEnc matrix. Recent tests show at such a high-res it works better than Andreas, and gives even more space savings. On a few small clips, I could bump the CQ 10 higher than normal and have it take the same amount of space (ie., from 50 to 60 with the kvcdx2 template).. it _should_ be similar if you are using the 2-cd kvcdx2 template.

[Martyn1980]

I dont have to compare the 2 side by side to notice a difference i dont even need the original to spot the difference i dont think, i can spot a DVD quality image right away if its from a proper film and not one of these TV Movie 4:3 things, i sit in front of Digital cable every night driving people mad by picking fault at the picture, telling everyone that i can do better than that, so i know im fussy, and all respect Injunpana but you thought that Matrix trailer was acceptable quality and i thought it was shit, and would go better to VCD than SVCD.

I crank the resolution on my monitor up to 1280x720 to spot every possible artifact and if im not happy then it wont go on a disc.

The reason i thought changing to Mpeg2 on Kwags stuff is i thought it might affect the final filesize a bit, and i dont have the time to mess about by trying it. I dont need the 3:2 pulldown most of the stuff i do is Pal and the NTSC stuff i do is usually 23.97 as this plays better on my DVD than 29 does.

I might try what you say with Gladiator, if i get time, cause i share this computer with someone else.
Is that the settings you want me to use and why not 10minutes and divide by ten its easier The problem i have with that is the first few minutes of a film aren't usually busy is it, if a film spends the first 10minutes with 2 people sitting talking then the filesize will be small will it not, this might not be an accurate way to do it, especially with action films at the end, i dont know that much about CQ on Tmpeg so correct me anyone if im wrong

[Daagar]

I crank the resolution on my monitor up to 1280x720 to spot every possible artifact and if im not happy then it wont go on a disc.

Well... here's part of the problem. If you are doing tests like this, almost nothing _but_ the original DVD will make you happy. Stretching a 480x480 encode up to 1280x720 is of course going to show artifiacts, even on a very well done encode. But you can't tell me those artifacts are visible on your 21" TV screen. Yes, I admit to seeing artifacts/blocks/etc. using Kwag's templates in VCD, but that's when I'm stretching a 704x480 clip to full screen (1280x1024). So long as I sit at least 10" away from my 25" TV, anything I saw on the monitor will not be visible now.

When people are _that_ fixed on the quality, why encode at all? Go get the original DVD!

[injunpana]

I might try what you say with Gladiator, if i get time, cause i share this computer with someone else.
Is that the settings you want me to use and why not 10minutes and divide by ten its easier The problem i have with that is the first few minutes of a film aren't usually busy is it, if a film spends the first 10minutes with 2 people sitting talking then the filesize will be small will it not, this might not be an accurate way to do it, especially with action films at the end, i dont know that much about CQ on Tmpeg so correct me anyone if im wrong

You must be one of those rare individuals who has good eyes. I don't, and even though I may spot a difference between a dvd and a compressed copy I don't let it detract from my enjoying the movie.

I buy dvds and just make vcds or svcds for a hobby. I know that in a year or two, dvd writers would become common and cheap. I am waiting for that because I want to collect movies in dvd format. In the meantime, I just make vcds or svcds out of the desire of wanting to find out if I can produce a near-dvd quality in 2 cdrs. And I give this to a friend or two if they want a copy, lol.

Yeah, that method of encoding the first 6 minutes and looking at its filesize is just a guesstimate. Obviously, the estimate would be more correct the longer the clip. Or you could just encode a 10 minute clip made up of fast action scenes, that would be best.

[Martyn1980]

I havn't really got very good eyes, cant see past 3 feet very well, but up close its perfect so i suppose thats why i notice more as im sitting so close to see it.
I mostly do this for a hobby just to challenge myself to see if i can do it the best i can, if i really want someone i buy it, i have 32 DVD's here where i am now and another 50 at my mums house, so i dont do this cause im too tight fisted to buy them.

I also dont try and get DVD quality on a computer at that resolution, i know i wont, but the closer i get on here at that resolution the better it will look on a TV.
To be honest i hardly ever watch much of the stuff i do, and sitting 6 feet away from my TV, i cant tell the difference between a CBR version of a film and a 4 pass VBR version, i just want to see how far i can push it within the limits, i have every intention of buying a DVD Burner in the next few months and then i imagine i will be trying to get 2 films on a single disc and as near to DVD as i can, or maybe getting films like Gladiator on a single layer single sided disc with no noticable drop in quality.

[vitualis]

Michael can't see what you are talking about. Looks like you just want to use CBR for everything - which is fine as it's what I do most of the time.

Obviously...

The point is that beyond a certain level, there is no benefit in using any VBR mode over CBR, if you are maintaining SVCD specs. This is because of (1) we live in a world of fixed media sizes and (2) SVCD specs has a ceiling bitrate.

For example, in this discussion about VBR encoding, there have been people talking about 4 disc SVCDs. In my mind, this is completely illogical unless these people aren't actually creating SVCD (e.g., XSVCDs) or they have a really long movie (something over 2hr 40min long).

As for the noisy signal problem, I stand by my assertion that a noisy signal is more difficult to encode and therefore a better test of an encoders abilities.

Again as before.

A test of an encoder's ability is its fidelity. Thus, a "perfect" encoder should be able to reproduce all the noise in the original. The issue is that this is difficult to judge with human eyes. We are good at picking noise from simple lines and shapes but are pretty bad at picking up differences in noise (encoding artifacts different from the orginal noise).

As explained before, an encoder could apparently do well in such an encoding test simply because it applies some sort of pre-encoding noise reduction or filtering. This, however, has nothing to do with the encoder itself and such techniques won't give rise to benefits from a good quality source.

Regards.

[kwag]

You can of course switch to mpeg2 for the kwag templates, though I'm not sure what it exactly does to the result. It should be equal to doing it in mpeg1, though. I personally prefer VCD, because my player requires the 'header trick' to play SVCDs. One nice thing with switching kwag's templates to mpeg2 is that you can turn on 3:2 pulldown.

Injunpana... try switching the Kwag template from the Andreas matrix back to the default TMPGEnc matrix. Recent tests show at such a high-res it works better than Andreas, and gives even more space savings. On a few small clips, I could bump the CQ 10 higher than normal and have it take the same amount of space (ie., from 50 to 60 with the kvcdx2 template).. it _should_ be similar if you are using the 2-cd kvcdx2 template.

All KVCD templates were switched back to "default TMPEG" Q matrix yesterday.
They are all updated at the download page here: http://www.kvcd.net/dvd-models.html

kwag

[energy80s]

[quote="vitualis"]For example, in this discussion about VBR encoding, there have been people talking about 4 disc SVCDs. In my mind, this is completely illogical unless these people aren't actually creating SVCD (e.g., XSVCDs) or they have a really long movie (something over 2hr 40min long).[quote]

Oh dear, since I do 4 disc CVD box sets I must be totally illogical, captain!

[vitualis]

If you are maintaing SVCD (or CVD) ceiling bitrates AND your movie is less than 2hr 40min, then it is damn illogical for you to be using any sort of VBR.

Regards.

[Martyn1980]

Thats a Lot of discs, i kinda see that as pointless, i only put lord of the rings on 3, i just like to keep them nice and tidy, most of my stuff is on 2 and i have the in 2 disc dvd cases with a nice printed cover so most people think all my dvd's are originals.

Virtualis there are a lot of cheap DVD players right now and most of them are from asia, so they tend to play they stuff we all do better than the big brand players, i think most people on here, probably do XSVCD more than anything cause you can get great results that way, ive even started doing VBR with a 300min and a 4000max bitrates now, and they play perfectly and i can see the difference easily, as my yukai plays nearly everything(high resolution Mpeg1 is a problem) but my Toshiba SD210E, played Swordfish perfectly, which had a max of 2520, but wouldn't play Driven at all and that was a max of 2600, although i have had 2900 working on it but i dont know how i managed it.

[vitualis]

Absolutely. I understand the benefits of higher max bitrates and it is a good point. The max. bitrate of SVCD/CVD of 2x CD is really quite limiting and if your player can handle it, you can get much better qualities but using a higher max. bitrate.

There are virtues with using the standard bitrates, however -- with regards to compatibility...

It's a bit of give and take.

Regards.

[int_53185]

HI all again! nice to see the discussion progressing so nicely after 10 pages! LOL i have a solution to the dilema...get a dvd burner! just ordered one today:P saying goodbye to svcd's...not looking back....have fun discussing this now mute subject! have fun!

[$$$]

>saying goodbye to svcd's

...and hello to CVDs the moment you want to have these three movies on one DVD-R. The topic ain't so mute as you might think now...

$

[bilbogod]

Oh well, feeling in a particularly contentious mood - I just thought I'd chuck in my observational spanner into this calming pool of lava

Having finished encoding Pitch Black, The Player, Coyote Ugly and Enigma with the KVCD (2 cd template) there is no way in hell I'm going back to CCE - 3pass within dvd2svcd. In scientific parlance, these encodes kick ass. OK so having to make the extra effort of learning how to use an AVI synth script combined with one or two tricks namely header changing saw a little patience and effort rewarded with my best results ever.....I particularly liked the way in which the four (2 minute) cq test encodes enabled a far more accurate determination of filesize than would otherwise be provided for by just doing a straight hit and miss cq encode.

Moral of the story - don't believe in idle hype and bloat.....just try it..you might quite like it

[Martyn1980]

Why aren't you going back to CCE 3 Pass, you never said why, Kwags Templates are improving and i quite like the new ones, but in my opinion nothing beats 4 or 5 pass VBR on 704x480/576 with CCE, can i have an explanation why your not going back, and if its about quality as it probably is, why is it better, and i dont want to hear about lackof blocks because lack of block noise isnt a measure of quality, like most people think after using Kwags templates.

I know i seem incredibly biased but it is only my opinion.

[bilbogod]

Actually I'm not gonna lie - I haven't run tests on all manner of combos to see which provides the most eye pleasing results (recently) since I no longer have the time that I used to. Yes my comments were quite lazy since they didn't justify my reasoning, but rather provided yet another dollop of subjectivity.
I was really targeting opinion towards the more point and click orientated newbie. Someone who doesn't spend all hours of the day trying to achieve visiual perfection, but would rather settle for a reuseable method that provides repetetively good results. I was particularly impressed by the way in which the dvd2svcd authors had dealt with the unpredictability of CQ by trying to ensure some level of consistency in total file size. Obviously this is not gonna be 100% perfect, but seems to be ok in most cases.
As for the visual look what can I say. Well the image appears to be optically brighter and sharper within the same previous constraints of 2 cds. Obviously increasing the resolution - subject to available space and bitrate will reap dividends and to this end I was extremely pleased with the results. This is on a comparison with your bulk standard 352 X 576 3pass+ with dvd2svcd using CCE. Where I found that a kind of heat haze effect albeit quite slight would still be perceptible from a viewing distance of 12 feet. I wasn't sure whether your suggestion above was based around Mpeg 1 or 2 ?? If it was mpeg 2 then I reckon we'd be talking 3 cds for some of the above movies. If it was mpeg 1 then the overwhelming opinion (yes I do get swayed by it) suggests that TMPGenc performs better on sub 2000 KBPS bitrates. For me there is an undisputable visual improvement which I would simply love CCE to be able to replicate given the speed of its encoding.........

oops sorry the most imortant thing that I noticed was just how much better things got when taking an extra couple of hours to use Fit cd in conjunction with the AVI synth script. Correct resizing and cropping is definately worth it and obviously allows for better allocation of bitrate. You get more bang for your buck.............and oh yeah........ I reckon providing a surround 2 audio stream courtesy of headac3he is another little delight that a few extra mins can provide you with

[Martyn1980]

I am kinda biased against Mpeg1 as my DVD wont play it ant any other resolution other than standard VCD, which is a bitch, and i agree that Tmpeg produces better results at low bitrates, but from my tests and what i like in a picture i'd say its more like 1500k rather than 2000, but its only a subjective opinion.

I use Tmpeg for anything with a noisy image, bad Divx stuff, because it show up less of the imperfections of the original, and ive recently started using Kwags templates with Mpeg2 instead of Mpeg1, and im getting quite good results, even though the quality is still well below what im used too, the filesizes are amazing.

I agree about Fit CD its a great tool and hardly gets a mention on here, except as a bitrate calculator when its so much more than that.

[bilbogod]

Hey have you ever tried muxing and remuxing with svcd headers using BBMPEG ?? That was the only thing that would allow my Eclipse 695 player to accept high res Mpeg 1's as well..........

[Martyn1980]

Thanks a lot, i'll give that a try