VBR vs Constant Quality

[mistermickster]

When encoding in TMPGEnc Plus, what is the difference between VBR and Constant Quality? Which gives the best results? I know you can determine the size of the file by setting the average bitrate for VBR, but how do you determine what CQ to use to get the size of file required?

Thnx.

[shoplifter]

You can't. :cry:
It does what it says: it gives (tries to) constant quality. For filesize estimation you have to try short samples of the movie i.e. 10 minutes and gues the final size. But there's no guaranty. Amd thats good, because than the constant quality feature wouldnt make sense. right?

[mistermickster]

But how do the final results of each method compare? Which gives the best output?

[aybesea]

But how do the final results of each method compare? Which gives the best output?

Given the same bitrate in both instances, VBR is far and away superior to CBR. My suggestion: choose the bitrate based on the length of the movie, then use VBR to encode.

[mmasw]

But how do the final results of each method compare? Which gives the best output?

Given the same bitrate in both instances, VBR is far and away superior to CBR. My suggestion: choose the bitrate based on the length of the movie, then use VBR to encode.

..."far and away"....??

that is "far and away" too generalized a statement to lend much value. For "high motion" video, VBR would certainly be better...although the visual advantages to VBR will diminish in equal amount as the quality of the source material degrades. For older analog sources like VHS, it is almost impossible to see a difference in the final video quality between VBR and CBR....at least I can't see much difference.

Accordingly then, encoding time becomes more of a consideration. VBR certainly takes much longer to encode then CBR. Perhaps, one would be better served using straight CBR, or even adding some filters rather then usingVBR....that is if you actually wanted to use your computer for other things then chugging away for hours (or days) on an encoding project.

[mistermickster]

Hi aybesea and mmasw,

I am talking Constant Quality not Constant BitRate!! Which I believe is not the same thing.

If I'm correct, Constant Quality uses a variable bitrate to attain a "Constant Quality" throughout the length of the movie.

My question is, how does this compare with using straight forward VBR?

Cheers

[mmasw]

Hi aybesea and mmasw,

I am talking Constant Quality not Constant BitRate!! Which I believe is not the same thing.

If I'm correct, Constant Quality uses a variable bitrate to attain a "Constant Quality" throughout the length of the movie.

My question is, how does this compare with using straight forward VBR?

Cheers

ahhh....yes, you said constant quality in your original post. Sorry I didn't pay attention. As to the differnce....I don't know about that. Have you tried typing "constant quality" in a search string here?....may be worth a try.

[ZippyP.]

CQ is a one pass VBR with an unpredictable file size, 2-pass can take twice as long but file size is certain. As to which is better, well it really depends on the settings used and your personal preferences. Video wise, I think the output can be very similar. I prefer 2-pass as I can fill the CD's and I do batch encoding overnight so time is not a factor. 8)

[poopyhead]

real-life experiences differ from one situation to the next...from one movie to the next....from one encoder to the next...from one source to the next...etc.

in theory, VBR totally kicks the crap outta CBR, which i find is mostly the case, especially if you wanna squeeze as much movie onto each CD-R as possible.

whether it be CQ or multi-pass VBR, VBR allocates a given bitrate much more efficiently than CBR. CBR gives each scene basically the same (not exactly the same, as i've heard...but close enough). So, this means that a slow scene, such as the end credits, would be given the same bitrate as a fast action scene, such as the matrix lobby shooting spree. makes sense?

if you would've used VBR, the encoder theoretically would transfer some unncessary bitrate from the end credits, which don't need very much bitrate, and give it to the fast action matrix lobby shooting spree, since it requires much more bitrate than the end credits. In the end, the fast action scenes will look much better, while the slow scenes don't look much worse.

as ZippyP. mentioned, CQ VBR is single pass. what it strives to do is maintain Constant Quality. this means that the same perceived quality for slow scene would require less bitrate to accomplish than a fast action scene. this is how bitrate is allocated. since this is only 1 pass, filesize will not be as predictable. however, encoding time is nearly the same as CBR....fast.

multi-pass VBR will go through the video several times, depending on how many passes you set the encoder to do. multi-pass VBR allows you to set an avg. bitrate. this avg. bitrate will be the bitrate of the entire video, on avg. this makes filesize much more predictable. for example, if several scenes are allocated bitrate that's less than the avg., then there will be corresponding scenes that will get bitrate higher than the avg. however, if you were to avg. the bitrate of all the scenes in the movie, it will be the avg. bitrate. however, since the encoder will be going through the video several times, encoding time can dramatically increase.

scenes=frames...but "scenes" sounds much better

[energy80s]

A lot depends on the encoder. With TMPG CQ is best, with CCE VBR is best. Oh and if you have the room (and its a short video) just CBR it at 8Mbps and you'll be fine.

[poopyhead]

With TMPG CQ is best

2 pass VBR is best, but nearly doubles the encode time...which may not be worth it, depending on the CPU.

[energy80s]

Not with TMPG!!!! Even the programme's author admits this.

[poopyhead]

Not with TMPG!!!! Even the programme's author admits this.

really?? didn't know that...is there a specific version that this is true, or is it in all versions of tmpgenc...

you got link where author said this???

btw, 2 pass VBR has much more predictable filesize than CQ VBR.

[energy80s]

Its common knowledge that TMPG doesn't do x-pass VBR well. It was optimised for CQ Coding. I agree that file size prediction with CQ mode is a pain, although I normally use a CQ of 85.

[poopyhead]

I normally use a CQ of 85.

wow...that's pretty high..on the border of becoming CBR....

normally, i've heard..and from personal experiences that CQ above 75 is overkill...just extra filesize with not much improvement in quality.

[EddyH]

If you have the extra disc space (eg short and/or low-motion movie) spare for a high CQ, you may as well use it and get super duper quality Overkill schmoverkill, so long as it fits. I just did a near-two hour film with very little action (hint: Spacey fruit) at CQ81 and was able to easily up the sound quality by two bitrate notches, there was so much space left from an 80min CDR... Looks great, natch.
(the SVCD maybe not so good, leveling out at about 63 even with kwag magic, if only the newest tmpgenc didnt have that durn memory hole that means i cant get more than 75% of it done)

For me, it's not a job, so I have plenty of time to kill, and tweak the settings up/down to fit inside a disc or get higher quality, so using CQ works pretty well. Better than TMPG's 2-pass mode anyhow, which doesn't vary the bitrate anywhere near as much. Even at low qualities (my insane obsession with 8cm films ) it keeps the movie at an even level of messed-upness that is easier to watch than one that may swing between not-so-messed and badly-messed with 2-pass (and worse with CBR); at high qualities it's not got the same fear as VBR of really opening the bit throttles on action scenes, or closing down to the minimum on the quietest parts.
But it isn't perfect - due to the nature of mpg encoding (or something ) it varies it just a little too much.. when the scene is quiet, your eye notices corruption more easily, and when things are whizzing around, less easily. A fusion between the two modes could be the perfect thing... or perhaps, a control similar to that supplied with XviD that can change the quality vs bitrate curve with more of a 'compression' style than simply closing in the max/min parameters. Apply it to VBR, voila, you can force it to go a bit more wild in it's variations rather than being CBR-with-knobs-on. Apply it to CQ, you can reign it in just a little.

Plus if you *are* in a hurry, not enough time for 2-pass, the unpredictibility of it may make you choose CBR and quality be damned instead.. kind of annoying that there's this handy mode and no way to even pre-scan a video file and estimate how much space it will take up. Going back to kwag and magic again, apparently there's a program (CQmatic?) available on the kvcd website now that does something between that and black voodoo to tell you quickly and almost exactly what CQ to use, but I dunno how it works or how long it takes.

[thayne]

I have always used 2-pass vbr with TMPGenc, but I decided to try CQ and the output was really good. I like it and its faster

[poopyhead]

If you have the extra disc space (eg short and/or low-motion movie) spare for a high CQ, you may as well use it and get super duper quality Overkill schmoverkill, so long as it fits.

heh....high CQ means CBR...you might as well go with that then...

IF there is any extra space...CQ should still NOT go above 75....what can be increased is the max bitrate.

when you increase the CQ too much, the encoder will start giving slow scenes more bitrate than it really needs. do you actually think scrolling text (i.e. end credits really needs that much bitrate)...well, the closer CQ gets to 100, the more it's like CBR where every frame basically gets the exact same bitrate.

HOWEVER, if you leave the CQ at 75, but increase the max bitrate, then the encoder will start pumping more bitrate to scenes that need it (i.e. fast action scenes) rather than wasting the "extra" space by pumping in more bitrate to slow scenes (i.e. end credits)

---------

for example, you're encoding the Matrix. at your current settings, you still have about 200 MB "empty" space left on the CD-Rs.

1) you could just jack up the CQ....which means that the bitrate of every scene will be more and more similar...(i.e. the difference in bitrate b/w slow and fast scenes will be smaller). so, with a jacked up CQ....tmpgenc will give each scene almost the same extra bitrate .... lets say 100 bit/s. this means the lobby shooting spree, AS WELL AS the end credits will receive this extra bitrate.

2) or you could leave the CQ alone and increase the max bitrate way up there. in this case, the lobby shooting spree may get 500 bit/s extra bitrate, while the end credits may not get any extra bitrate at all.

now both methods will fill up that "empty" space, but one does it more efficiently than the other. take a guess which one

[jolo]

The statement that aybesea said about variable is just not true.
It s personal choice amd both have pros and cons.

Variable rate will tend to pack in high quality while using less disk space.
CQ, CBR will create very high quality, with more disk space and take less time to encode. CQ or CBR also opens up a lot more tools and possibilities for editing and spillting the document afterward, while variable rate can be very tricky and not many editing applications can work with it.

I usually work with VBR and encode with Main Concept which will split my files for me. But if I plan to do editing with it, I use CBR at the highest bit rate I can do.

[poopyhead]

@jolo,

1) CQ is a form of VBR...so it is still much better than CBR and should NOT be grouped w/ CBR.

2) since we're talking about (S)VCDs here, quality:filesize ratio is of utmost importance

CBR is the least efficient in allocating bitrate and should only be used on high capacity media, i.e. DVD. but even with DVDs, if you wanna fit more on each DVD-R/+R, VBR allows you to fit more per disc w/o losing too much quality.

[EddyH]

but my min/max bitrates are already at the limits of what my player can handle (about 64/2560kbit for mpg1, with 160/192?k audio). thats the beauty of the thing maximum flexibility.

as it stands i might raise the minimum as the quiet scenes are maybe getting not quite enough bits (gibbs/mosquito noise on edges, faintly) even at CQ81 even though a few scenes are pegging the yoke up at 2560 (and looking mighty fine compared to what the disc's CBR equivalent - 790kbit - did in testing, might i say)



Jolo just reminded me of a small problem i'm having what with the program bombing out 75% thru - CQ streams can be tempermental when trying to join them, esp with TMPGenc (refuses to do it often as not). Does anyone know of a good alternate mpg splitter/joiner that would be a little more willing to even try joining the two bits?

[energy80s]

Trust me, a CQ of 85 ISN'T the same as CBR! I know as I use both! As I have my max bitrate at 8000, upping it isn't an option. Sometimes I use lower values down to 70, but there is a very noticable difference in bitrates at these lower CQ values, and I don't want my encodes to suffer - rather splitting across two discs Likewise I don't like low bitrate fullsize MPEG2, so I don't like the bitrate dropping below 3000, most of the time I keep the low ceiling at 4000. Multipass VBR with TMPG isn't the best option. In fact it's the worst of the 3 main options.

[poopyhead]

Trust me, a CQ of 85 ISN'T the same as CBR!

never said that.... i said as the CQ gets higher and higher, it is nearing CBR set to max bitrate...

CQ=100 is nearly the same as CBR at max bitrate.

As I have my max bitrate at 8000, upping it isn't an option.

we're referring to (S)VCDs here...so bitrates aren't quite that high yet

Multipass VBR with TMPG isn't the best option. In fact it's the worst of the 3 main options.

yea..i've heard recently that tmpgenc's 2 pass VBR isn't that good....someone said tmpgenc was designed more for CQ than 2 pass...which is strange...

i guess only multi-pass on CCE then.

-------------

@EddyH,

i would definitely stick with 224 kbit/s audio....lowering bitrate generally won't save you very much space, but the quality will be noticeable.

but, yea...i think it's better to adjust the min/max bitrates than the CQ...CQ has to deal with bitrate allocation...and you want to set that as efficiently as possible....changing bitrates would prolly be a better solution to fill up "empty" space

[lordsmurf]

So, this means that a slow scene, such as the end credits, would be given the same bitrate as a fast action scene, such as the matrix lobby shooting spree. makes sense?

That's a bit misleading. The bitrate isn't dependent on later/earlier scenes. It allocates what is needed for what's on screen at the time.

First you specify the average, then give it leeway for min and max. When you have high action, it tends to use the higher since it was allowed, like your high-action. When it's people standing or talking, then a lower bitrate may be used. It's a case-by-case (or screen-by-screen) basis. The first pass just analyzes how much each image needs without corruption, then applies in the second pass. The first pass also takes temporal information into account. The second pass (which is nearly identical to the first pass on 1-pass VBR), applies the compression from the info it is given (and the 2-pass method has more info available for this determination). The first pass gives temporal information, whereas the second pass does not (much like the 1-pass method). But the temporals are for smooth movement and macroblock prevention, not for "taking" or "giving" bitrate to another sector of the video footage. The 1-pass vs 2-pass is more like prediction vs hard knowledge. The 1-pass has to predict what's next. The 2-pass knows. More passes is retarded, and serves no real function other than to waste your time.

End credits actually DO take up a decent amount of bitrate due to continous movement, so that's a bad example, especially if the credits are not just white text on a black background.

VBR is only good for fitting on media. CBR and VBR will look the same at the high points. If you use 8000 CBR, it'll be just as good as VBR with a max of 8000. But with a medium CBR of 5000, then a VBR of 5000 average with max 8000 may look better. MAY! Likely, but not always true. This is for 720x480 MPEG2. Half those bitrate numbers for 352x480.

CQ is a variant of VBR, and looks better than 1-pass VBR in TMPGEnc. The 2-pass vs CQ argument is as useless as the TMPGEnc vs CCE argument. The differences are minimal and the "better" arguemnts are only based upon source and user opinions, and nothing more. Both look great, if done properly.

An encoder that supports multipass will in a first pass analyze the video stream to be encoded and write down a log about everything it encounters. Let's assume that we have a short clip which starts out in a dialog scene where we have few cuts and the camera stays static. Then it leads over to a karate fight with lots of fast cuts and a lot of action (people flying through the air, kicking, punching, etc). In regular CBR encoding every second gets more or less the same bitrate (it's hard to stay 100% CBR but that's a detail) whereas in multipass VBR mode the encoder will use the bitrate according to its knowledge about the video stream, that is the dialog part gets of the available bitrate and the fighting part gets more bitrate. The more passes the more refined will the bitrate distribution be. In single pass VBR the encoder has to base his decisions on where to use how much bitrate solely on the knowledge of the stuff it previously has encoded.

However, with all this being said, some programs, using the "bitrate calculator" functions, can specify the desired end size in VBR, such as the Wizard used in TMPGEnc. In this case, the bitrates are more rigidly controlled, and the temporal pass tried to take into account the maximum file size you have specified when determing what it shall allow on the scenes. This is not the best method of VBR, though the multipass becomes a necessity at this point. It is with this function that it sort of compares footage as poopyhead suggested, but it is still not stealing/giving bitrate from one scene to another. In fact, with a specified file size, if it is smaller than what the uncontrolled VBR would had done, it'll just steal bitrate from everything. It affects bitrate too.

CBR vs VBR vs CQ, especially tossing in forced disc-size restrictions, if complex. Hopefully I didn't make any of it confusing or misword anything, but this isn't an easy topic.

I tend to use 2-pass or CQ or CBR in TMPGEnc, depending on the project. I think CBR is best in TMPGEnc. I use CBR in Procoder. I capture MPEG2 in VBR (single pass), and it looks great, though you can sometime see temporal blocking at harsh changes in scene (would also appear in CBR). In all cases, VBR is just for file size.

[poopyhead]

So, this means that a slow scene, such as the end credits, would be given the same bitrate as a fast action scene, such as the matrix lobby shooting spree. makes sense?

That's a bit misleading. The bitrate isn't dependent on later/earlier scenes. It allocates what is needed for what's on screen at the time.

i was referring to CBR, not VBR, in this statement.

CBR only deviates slightly from the set bitrate.

------

if you're correcting my VBR analogy..i was just simplifying the explanation.

especially for multi-pass VBR....the encoder does have a time element as it needs to satisfy the avg. bitrate.

--------

btw, end credits obviously do require less bitrate...even though the text is moving....there is only a small amount of pixels on the screen that isn't black in any one frame during the end credits....compare that to any other scene and you will get larger variations between the pixels, requring more bitrate.

[gmatov]

End credits actually DO take up a decent amount of bitrate due to continous movement, so that's a bad example, especially if the credits are not just white text on a black background.

Lordsmurf,

I'm happy to see that in a post. The opposite has been written a hundred times.

The only convert I have to verify it is a Potter movie, the first 3 CDs of which have the movie and some of the credits, the 4th disk is over 50 megs of credits only. I don't know exactly how much time that encompasses, but I seem to recall the credits run a fair time. Recall tho' some are on disk 3.

I'll have to check the time, I'm curious.

Cheers,

George

[next]

Not trying to flame here.

CBR always worked for me. It was fast.

Loved it. And for the most part it was SVCD.

[next]

Is anybody interested in the real world as opposed to theory?

[lordsmurf]

End credits actually DO take up a decent amount of bitrate due to continous movement, so that's a bad example, especially if the credits are not just white text on a black background.

Lordsmurf,
I'm happy to see that in a post. The opposite has been written a hundred times. The only convert I have to verify it is a Potter movie, the first 3 CDs of which have the movie and some of the credits, the 4th disk is over 50 megs of credits only. I don't know exactly how much time that encompasses, but I seem to recall the credits run a fair time. Recall tho' some are on disk 3. I'll have to check the time, I'm curious. Cheers, George

MPEG bitrate is based solely upon algorithms to maintan fluidity of movement and/or retain image quality, with VBR asking what any lazy person does: what's the least I can give and still get decent results. More movement equals higher bitrate. A full-color scene that does not move can actually use less bitrate than the continous scroll of simplistic end credits. In VBR, the first image needs the most, then slacks off the longer the same data is on screen. On credits, new information appears in a constant stream, so it cannot slack off as much as a still of a person talking, where the only thing moving is lips. I have interview tapes where the footage I shot was less size than the end credits. Anything that stays on screen for an extended period (several frames, which can be less than a second) will use less bitrate. Any motion, and it's over.

[EddyH]

@ poopy

well.. umm.. i've used 224kbits before, sure, but only very infrequently. mainly for music VCDs where the sound quality has to be foremost, and keeping to standards is prefereable too. My preferred encoder at the moment is almost indistinguishable from CD at 224k except on a few 'problem sounds' (like most encoders have), both perceptually and technically (soundgraph etc). Lower qualities i have little trouble with, if they're filtered right. It does get a bit ropey from 160k down, but if i sacrifice some treble it comes out acceptably. 128k with a bit of care sounds like 224k in old TMPGenc and i even used 112k on a movie encode where the source Mp3 wasn't so hot - and space was tight.

160 (with filtering to a little better than 128k mp3 equivalent) or 192k (without bothering much to filter) usually sound damn good, certainly enough for playing through a TV and/or a modest stereo (when i get something that will make it sound horrid, i'll be able to afford the actual DVDs, and/or a DVD burner). The current state of most mp2 encoders is a little slack, either poor output, really slow, or both, but im sure ive pounded on too much about that already... and im not a programmer or audio/maths graduate so cant rightfully bitch.

------

@ next: word. it's all about your own experiences and comparing them to other peoples, in the end.