Cbr - Vbr question

[{morpheus}]

Hi I'm pretty new to the svcd scene so please be tolerent.

I have done a lot of testing using cbr & vbr recently.

apart from:
1: vbr taking a hell of a lot longer to encode than cbr
2: a slightly bigger file at the end.

what else does variable bit rate encoding give you?
does it give better quality? if so in what area's?

I tend to use cbr 2150 for all my movies & they look great!
& they only take around 7/10ths of the original movie time to encode.

am I missing something? :confused:

[ScooterMyth]

Nope....as far as I'm concerned I have found that CBR for MOST (not all) movies is just fine. Most of my encodes are 1900-2100 CBR and they look fine to these tired old eyes. I get real time encodes or better with CCE and the time saved is tremendous versus 4 pass VBR. It's purely subjective.

[jdobbs]

It depends on how much space you have and what you're doing. Since SVCD has a bitrate limit (2X transfer speed]) the BEST picture you could get would be at a constant bitrate (CBR) at the maximum.

But... that would also limit the amount of video you could get on a disc. For example, if you encode at 2350Kbs video and 128Kbs audio, you could get about 45 minutes (rounding) on an 80min CD. While your video is playing, though, there are periods (like during still motion scenes or blank areas), that could have a level of quality that is just as good with a video bitrate of 200Kbs. That's a lot of wasted bits!

What VBR does is allow you to reallocate the extra bits in low demand parts of your video to the higher demand parts (where the they are needed). That way you could get (in the example) 2350Kbs (MAXIMUM bitrate) for the hard parts, low bitrate (MINIMUM bitrate) for the easy parts, and space left for more video! The AVERAGE bitrate setting strikes a balance and tells the encoder what your size goal is.

So it comes down to this: The BEST picture in SVCD is at the maximum bitrate. No one can argue that (assuming the same encoder). But by using VBR you can sometimes fit more video on a disc -- with no loss in quality.

I would recommend constant bitrates at the max for your home videos (where you're not trying to get a movie on 2-3 CDs) and use VBR for creating most movies. On the other hand, if you have an 80 minute movie -- you might as well use the max bitrate and be done with it (on 2 CDs)?

[Airw0lf]

Can anyone explain the benefits of robshot's VBR method over the simpler way of just hitting multipass vbr, and entering the max, min and average values straight away?
Because from my point of view, I can't see any gain in manually increasing the average bitrate in the advanced options because if you can get a low Q factor with a low average bitrate, you might as well just increase it to the maximum amount your available space will allow. And in some cases, even when you've reached your filesize limit, you will still have high Q levels, but there's not much you can do about it. Even if you decided to use another CD, you'd soon end up being hit by the 2600 kbit video bitrate. (And of course, we mustn't forget audio considerations as well!)

[mb1]

robshots method is not that good ...

Best way is to lower encoded macroblocks especially if 2.600 max. bitrate or similar isn´t enough any longer.

To do this you have to consider tv-overscan-area and the cce-bug.

Simpliest way to achieve correct values is to use FitCD
http://members.tripod.de/fitcd/

explained that in several threads ... but nobody understands that ;)

[Airw0lf]

which CCE bug are you referring to?

[jdobbs]

The real advantage to robshot's method is that it gets the the file size down to the minimum size possible while maintaining an acceptable quality level. When you do multiple pass encoding the size of your file will always be what you put in your AVERAGE bitrate. You'd have to encode it multiple times and look at the output to see how acceptable it is at that bitrate. With the robshot way you can use the ADVANCED tab (after only one pass) to get a prediction on the quality -- and set the bitrate to the lowest acceptable level.

When you're making SVCDs it really doesn't matter. You set the bitrate at the highest level that will fit on the number of CDs you choose and live with the quality you get (most of the time good). But when you're trying to stuff multiple movies on a single DVD-R robshot's method minimizes the output filesize and helps you take better advantage of your DVD-R real estate.

I just ignored the ADVANCED tab (I made mostly SVCDs) until I got a DVD-R burner and read robshot's article -- then it hit me, there is some real useful functions in there! My hat is definitely off to robshot.

[Airw0lf]

jdobbs, yeah I agree with you there. You can get the most out of the method with DVD, not really with SVCD.

[mb1]

Read FitCDs readme ...

"...CCE has a bug with the motion-vectors exceeding the resolution-boundaries.
e.g. if you want macroblock-optimization and a destination-size
of 448x320, witch is exactly 560 macroblocks: resize to 448x318,
because otherwise CCE would code the black macroblocks above and
below your film-resolution, too. 448x352=616 macroblocks would be
coded."

Unfortunately the site which shows this bug grafically is down at the moment.


From DVD-Source you can make quite good SVCDs optimizing first tv-overscan and macroblocks (FitCD) then use an optimized Matrix (f.e. Angels 1-CD-Matrix which is also good for 2 or 3 SVCDs) and then (if you want) using robshots method.
This gives the best results (and even if you don´t do robshots method you get better results than if you only use robshots optimizations).
You simply don´t encode picture-area which is unvisible on tv-sets (tv-overscan) and therefore get much better quantization ...

[{morpheus}]

Thanks everyone!
I am certainly a wiser man :eek: