Easy Question for You pros CBR/VBR ??

[alsyed]

i sometimes have to fit more than 3 hours on a dvd-5. If the original footage was excellent quality, then for me, on my 29 inch crt, the results are still quite good.

I calculate by hand what bitrate i will need, starting off by the size of the dvd.

for example, a stereo dvd,

dvd size: 4,592,762 KBytes
durations (3 hours) = 180min = 10800 seconds
audio size = 192Kb/s / 8 * 10800 = 259200 Kbytes
muxing/menu size (150MB) = 153600 Kbytes

4592762 - 259200 - 153600 = 4,179,962 Kbytes

divide by 10800 sec = 387Kbytes/sec = 3096 Kbits

which is your average bitrate to fill the disc.

i set my minimum to 300, and my maximum to be around twice that of the average (obviously that's limitied to around 9Mbps) and do 2pass vbr on that.

Its a simple 60 second calculation, that works for me.

If i am doing small vid clips, then i use CQ. I only use 2 pass vbr if i need to specify output size with the best possible picture for that size and duration.

[jimmalenko]

@vhelp:

You know your shit, and it is only your opinion after all, but I have a rebuttal or two to make:

These days, I prefer CBR at 9000 bitrates. I'm refining my process still. But, CQ offers a lot as well, to those that don't want CBR
in their process.

Some encodes work better w/ different MODE settings than others.
For instance, CBR does better on Interlace-only sources (such as CAM
footage) vs. VBR. Because (IMHO) VBR does not help enough
when the scene is very fast. And, I've come to believe that as long
as you think in tirms of 60 minutes to a DVD (home footage) that's
the goal. And, CBR at 9000 seems to closely make it. (I'm having
trouble hitting that mark - fitting 60 min on a DVD disk w/ CBR, but
I suspect I'll figure it out) .. plus, w/ CBR (and home footage) you
will not have to wait those long hours.

According to the above, unless you're using PCM audio, then you should have no problems whatsoever fitting a 60 minute 9000 CBR encode onto 1 DVD.

Because CQ is not CBR, there is a *greater* chance of fitting a 1 hour tv program/CAM footage on a single DVD disk. Yes. Anyways.

I think *greater* is very debatable. With CBR, plug the time into the bitrate calculator, use that bitrate and it will fit, simple as that.

***

Ok. Lets say, you have a 90 ~ 120 min movie, but you don't want to wait the hours of multi-pass encode (tipically 2 pass) and need the speed (as we all do) and *quality* thereof..
.
You could do it w/ 2-pass (w/ semi gaurantee of quality) *or* you could
go the CQ route, and go for broke.. because your final destination is DVD
..but half the time. CQ offers that.
.
You don't have to worry about space, fitting on a DVD disk, because the
chances of a 2 hour movie (using CQ) not fitting on a single dvd disk is
almost nil.

IMO if you choose the appropriate frame size and bitrates, quality with 2-Pass VBR is not an issue. I see it as a far greater risk that CQ will go over 1 disc, as 2 hours on 1 DVDR is an average of 4800kbps or thereabouts .... that will make any predominantly fast-motion footage be massive because it will be maxing out the bitrate the majority of the time.

***

If you use VBR (2-pass) you're gonna be waiting.

Waiting, yes ..... but waiting in the knowledge that it will fit 1 DVD, plain and simple. One encode, no need to "fudge" settings, nothing, nada. Right first time

***

Lets say, you want to capture and encode a 60 minute movie. You might get away with it using CBR and 9000 as your bitrate (though I have doubts that it will succeed and fit on a single DVD) ..

Proven that it will above.

[BSR]

Vhelp,
LOL, every time I read one of your posts I have to make sure I have a full beer because I know I'm going to be there for awhile.
Great post by the way.

All 3, 2pass VBR, CBR, CQ have their place.

I use 2 pass VBR for pretty much everything so far because I want to fit things to a certain size, usually the size of a DVD.

I'm not a CBR fan because I don't like wasting space, i.e. bitrate on scenes that don't need it.

I have a bunch of home VHS videos from the late 80's and early 90's. This is when I can see CQ coming in handy. I don't care how many DVD's it takes to hold all the footage, but I do want great quality. So I'll encode the tapes in the order of the year, do a CQ encode and fit what I can per disk, and go on from there.
Thanks Vhelp, I'll be giving it a go.

joecav - love the helmet. I thought this was going to be a rebuilding year for us but luckily to my surprise they have exceeded expectations. Gotta love it. Just hope Ben keeps up the good work.

[bugster]

@vhelp, I have read a lot of your posts, and you pretty much always hit the nail on the head, but in this case you are wrong, and I am surprised.

If you can't fit 60 mins of video on a DVDr using CBR at 9000kbits either you are using PCM audio or you are doing something seriously wrong.

* jimmalenko wrote:
Quote:
As bugster said, if hitting a target size is important (and it is to me ) then CQ is not really appropriate.


Not so. (read below for in-depth) ...

In short.. a one hour program (be it TV or CAM footage) and CQ, WILL
fit on a DVD disk Really !!!

There you are right, one hour WILL fit on a DVD disk, no suprises there.

]Because CQ is not CBR, there is a *greater* chance of fitting a 1 hour
tv program/CAM footage on a single DVD disk.

CQ, with the max set to 9000, will always produce a smaller file than the same video encoded with CBR 9000. Why?, because some scenes will use less than the max, even with your quality set to 100, and so the average bitrate will be reduced. It is the average bitrate that determines the final filesize, not the max.

I will always agree that TmpGenc's CQ mode is as good as 2-pass VBR when it comes to the final quality. The drawback is that you have no way of accuratley predicting the final filesize. So if your movie is long (> 2hours say) the chances of CQ producing too big a file to fit on a DVDr are to high for my liking. In the same vein, the chances of using too low a CQ quality setting and not maxing out the quality available are also too high for my liking. So I use 2-pass VBR. Sure it takes longer, but at least I don't have to do it again because I got it wrong!

Some will say if its too big, just author anyway and use DVDshrink or similiar to make it fit. I say get it right 1st time!

[jimmalenko]

No surprise that this thread is generating as many different opinions as there are posts It's like asking "what's the best car to buy?"

Capmaster *did* hit the nail on the head there, didn't he ?

[joecav]

joecav - love the helmet. I thought this was going to be a rebuilding year for us but luckily to my surprise they have exceeded expectations. Gotta love it. Just hope Ben keeps up the good work.

Here We GO!!
Hehe...

[LSchafroth]

joecav - love the helmet. I thought this was going to be a rebuilding year for us but luckily to my surprise they have exceeded expectations. Gotta love it. Just hope Ben keeps up the good work.

Amen! Go steelers! Just think, Ben is wearing Black and gold now. The last time he lost a game was to a team that had to get permission from teh Steelers to change their uniforms to a similar Black and gold. (can you tell I'm a hawkeye fan?)

Let's hope then can keep home field advantage in the playoffs.

LS

[rhegedus]

If I could be permitted to kick start this thread again......

You might find this link helpful in deciding CBR/VBR bitrates:

How to set optimal bitrate for MPEG

What also has to be considered is the source. Home cam recordings will need a higher average bitrate compared to DVD conversions to look good because all that camera shake just chews up bitrate like no tomorrow. Anything less than 6000kbps for home cam stuff should go the way of VBR.

As a rough rule, for VBR I've found that min and max bitrates -/+ about 25% of average work well enough.

[daamon]

As a rough rule, for VBR I've found that min and max bitrates -/+ about 25% of average work well enough.

By following this, don't you run the risk of throttling the encoder in it's VBR mode?

I mean, let's say an AVG comes out at 6,000. That means your MAX and MIN will be (around) 7,500 and 4,500 respectively. With lower averages, you get towards (to me) a preferred lower MIN, but have a greatly lowered MAX, and vice-versa for higher averages.

The max that low doesn't allow the encoder to utilise all available bitrate as far as the spec is concerned. Surely, something like 8,500 is better (depending on audio bitrate of course)?

And a min that high means that the encoder is restricted from going really low when it can - meaning that, to meet the average, high bitrate scenes aren't going to be as high as they could be (on top of having a "reduced" max) and low bitrate requirement scenes are gonna have too high a bitrate unneccessarily.

I'm not saying you're not right, but (to me) that rough rule seems just wrong. I've always followed the "have a high max (say 8,000 - 8,500), low min (say 0 - 1,000 - 2,000) and take the average from a bitrate calculator" rule.

If I'm missing something, can you please explain your rationale? Ta.

[adam]

That's my opinion as well daamon. If you are just trying to prevent the encoder from allocating too little bitrate to certain scenes, then adjust the bias or in the case of TMPGenc, the P picture spoilage. Arbitrarly adjusting min and max is an inefficient way to handle these problem scenes, because it handicaps the entire encode.

I always use 0 for my min and (10.08Mbits - audio bitrate - overhead) for my max, and then adjust my bias according to whatever average bitrate I am using.

[SatStorm]

TMPGenc 2.5 (NOT the latest version, I don't have time to test it those days), has 2 - 3 issues to solve, that make it "unique" the way it works...

For example, it does a bad resizing (only bilinear) so when you convert any source (except 704 x 576/ 480) to 352 x 576/480 (AKA 1/2 D1), you have distortions in the picture, especially on lines... This is very obvious on animation stuff. The solution is to resize outside the encoder (I do it with virtualdub thourgh frameserving).

Another issue it has, is the way it spread bitrate per frame / field. On static scenes, and when you use any VBR mode (including CQ & CQ_VBR), it goes the lower possible and that creates macroblocking for no reason! The solution is to rise the bitrate to a point that this effect disappears. IMO that is ~2300 kb/s which is very high for a minimum vbr value, but still enough to have about 4 hours per DVD5 using 1/2 D1 (just use CQ 2100min - 6000 max and you gonna see...)

A third issue is the YUV> RGB > YUV convertion it does, because of the use of vfapi plug in.... This "problem" shows more on NTSC than PAL, and it seems that sometimes, it can make bad things pop up during the encoding... A higher minimum bitrate helps...

Those things count on TMPGenc plus. The other encoders do a "better" (the correct term is "different") job: Mainconcept don't have issues with the minimum bitrate, so procoder (an encoder that I know only a bit...).
CCE does an excellent bitrate allocation per frame, so in a way it is the best choice for multipass VBR and lower minimum bitrates (but not a wise choice for CQ / 1 pass VBR & lower values)

Software realtime mpeg 2 is another bitch to handle... With mainconcept, I use min 2000 average 3000 maximum 5000 kb/s and I really like the results (1/2 D1). When I set the minimum lower than 1700, the result - in my eyes - suffers...
With ATI ask Lord Snumf, he is the expert
With powerVCR 3, a minimum of 1000 is okey. After all, the fucked up way this software works, don't make sence the first play (it just works... when you manage to make it work - and it works good when that finally happens...)
WinDVR don't care for anything: use lower bitrates and any framesize you output is a product of 352 x 288/240. Use higher bitrates and the output is a product of 2 352 x 288/240 frames, in an offset position between them and blended picture...

There are some other encoders out there too, but I never test them to much.

Just my 2 ? cent

[adam]

Satstorm, if you find that the quality of some still motion scenes are suffering during drastic dips in bitrate when using TMPGenc, then you need to lower either your P picture spoilage or b picture spoilage settings, or both. That is what this setting is there for. I don't think arbitrarily raising your min setting is an efficient way to handle this issue.

[rhegedus]

The average -/+ 25% for min and max bitrates is just a rough guide that I’ve found works for me.

As there's nothing to be gained by increasing bitrate over 8000kbps for a CBR encode, I figure that a max of 7500kbps using VBR would produce the same results (since it is believed that VBR encoding also does more than allocate bitrate better).

What we also have to consider is the video time that an encoder can spend at max and min bitrates while achieving an average. It's all very well trying to eek out as much as you can by setting high max bitrates, but there's no give without take: the higher you set your max from average, the more frames will need to be encoded at lower than average bitrate.

For example, if you set a min of 1000, average of 4000 and max of 8000, for every 3 frames encoded at max bitrate, 4 will have to be encoded at minimum bitrate to produce an average of 4000. Of course, there are a wide variety of bitrates in between that could be used or even more frames at a slightly higher than minimum bitrate but you get the drift.

The closer to maximum the average is, the more even the distribution of bitrate between frames, but if were using high averages to begin with, there probably isn’t much benefit in going the VBR route.

I agree that with footage which has a lot of still scenes does lend itself to a wide spread of min and max bitrates, but most of the stuff I encode tends to be quite bitrate hungry and so a narrower spread of bitrates with a reasonably high average suites my purposes well.

If you are just trying to prevent the encoder from allocating too little bitrate to certain scenes, then adjust the bias or in the case of TMPGenc, the P picture spoilage.

Learning something every day: I’ve just noticed this in the manual VBR tab – how does it work?

[SatStorm]

@adam: Unfortunatelly, TMPGenc had bugs and issues, that made theory and practice uncompatible! TMPGenc use to be a special case, as we have discussed various times over the years...
The latest TMPGenc 2.5 version, has some interesting bug fixes...

Let me copy / paste some of them (from our own site here: https://www.videohelp.com/tools?tool=TMPGEnc_Plus

Specification Changes
[4] [MPEG Setting - GOP structure]: Even if Closed GOP is not selected, the GOP structure is now compliant with the MPEG standard.

A really - really OLD issue, finally solved...

Image Quality
[2] Encoding of scenes using still-image, has been improved.

That is the problem I said: It was a bug. It was always a ******* bug... Sorry about the expression, but if I count all those hours I spend over the years to eliminate this issue, I would end up with a number of months that I could do much more things in my life, than testing a ******* bug of an encoder...

Decoding
[2] The encoding of MPEGs with GOP structure using only I and B pictures, has been improved

Does that mean that the GOP structure works as the theory determines? I hope so!

[3] [MPEG Setting - Video]: The Aspect ratio 2.21:1 for MPEG-2, was set to 2.11:1 .

I hope that means "bye bye anti alizing!"... I expect my 352 x 576 projects look far better now, even if this bug fix is not direct connected with that issue

Unfortunatelly, I don't have time to test in detail all this things yet... But I hope the best!

[midnightsun]

rhegedus, setting a min and a max in VBR just means that the encoder CAN go as high as max or as low as min, but doesnt necessarily do so. It does if it finds the circumstances call for that. What others were trying to say I believe, is that if you set your min and max close to each other then you're limiting the encoder potential to allocate bits where needed and take them where "redundant". In that case there's no use in 2-pass VBR but 1-pass CBR would be about the same quality-wise and, not to be overlooked, twice as fast.

[rhegedus]

I realise what your saying about max and min bitrates, hence my 'wide variety of bitrates in between' comment.

I don't necessarily think that having a 'narrow' spread limits the encoder, after all using -/+ 25% means the max bitrate is nearly 70% more than the minimum. After all, just how mush leeway is there for VBR encoding of VCDs and SVCDs?

Granted, DVD bitrates are huge in comparison, but the obsession with huge max bitrates in pursuid of quality seems just as wasted when there is no benefit over a CBR 8000kbps encode.

[adam]

Believe me SatStorm I disklike all those bugs as much as you. What I am saying is that there are better ways to work around some of them then just raising your min and lowering your max. I cannot think of any reason why you would ever want to do this. I cannot see how it could possibly help quality any, but I can surely understand how it could degrade it.

rhegedus, once you start using high bitrates these types of things start to become less and less important. If you are unconcerned about bitrate distribution because of your high average then just use CBR. If you are going to use VBR then just use the widest reasonable interval from which to select your bitrate. There is no reason to limit it, it accomplishes nothing and yes it most certainly will lower quality. Maybe doing so will still make your bitrate allocation "good enough" but maybe not. Why chance it?

BTW the B and P picture spoilage settings should be present in more then just the manual VBR mode. In my version they are there for manual, CQ, and 2-pass (new style not old style VBR). The number you use represents the % that, that picture type can be reduced in quality.

[mpack]

IMHO, CBR is obsolete. It died with VCD. Enter intelligent encoding and VBR. Why run your bitrate up high enough to not get macro blocks in flashing or fire scenes, and use the same screaming high bitrate for two talking heads sitting still at a table?

I don't really agree with that. On VHS captures for example my experience is that no amount of passes with a VBR encoder will make the video look good, so you might as well do a quick and dirty one pass CBR encode at the max bitrate you have room for. Of course, CQ is just as quick, but as others have pointed out you can't target a particular disk capacity with that.

[BSR]

IMHO, CBR is obsolete. It died with VCD. Enter intelligent encoding and VBR. Why run your bitrate up high enough to not get macro blocks in flashing or fire scenes, and use the same screaming high bitrate for two talking heads sitting still at a table?

I don't really agree with that. On VHS captures for example my experience is that no amount of passes with a VBR encoder will make the video look good, so you might as well do a quick and dirty one pass CBR encode at the max bitrate you have room for. Of course, CQ is just as quick, but as others have pointed out you can't target a particular disk capacity with that.

I think what he's saying is use a 1 pass VBR. You would theoretically get the same quality with room to spare(smaller size) because not all scenes will require the full bitrate.