Best way to determine average bitrate for Svcd? (If TMPGenc had a brain?)

[Brother Darrell]

I don't want the encode to be constrained to what I THINK the average bitrate to be, I want to find out what it would be if it had free reign to determine the average bitrate it's self.

I realize this is based on size and time.I also realize Tmpgenc has a few rate control modes that are capable of finding it's own average bitrate.
If I want NTSC FILM, 480x480, audio 192k, and I don't care how big it gets, just as long as it is not bitrate stuffing; Then would AutoVbr at 100 be better than ConstQuality set at 100?

[edit] sorry, felt I needed to clarify what I am really asking.

This DOES sound like one of those posts that is asking "WHAT IS THE BEST BITRATE FOR"...yada yada yada.

What got me thinking about this was the average quality slider in ConstantQuality. No matter what I set the maximum bitrate as, if the slider is at 100 then it will tend to stay around the max bitrate. I know from experience that an svcd video does NOT need a max bitrate of 9000. But, for grins, if I set it there , it adjusts all other frames accordingly. I know it does NOT need all those bits, it's stuffing the bitrate & I end up with a needlessly oversized file.
I have certain settings I like to use that work very well. I generally set my max at 2600. I don't SEE anything making this into an under-nourished encode,They all end up looking pretty damn good. What's the beef? I can't help but feel it's still stuffing bits into scenes that may not need it. Then of course there is 2pass that distributes the bitrate where it needs to be, but depending on what average I choose, it will stuff bits too. Don't get me wrong, I LIKE having that control,And it's needed for final size control. But you would think (at least I do) there would be some way to tweak just a little bit more, to use the least amount of bitrate needed to aquire a desired quality. You can do this with CCE, and I thought there should be a way to approximate this kind of tweaking in TmpgEnc.

[davexnet]

I've heard it mentioned that CQ does provide a bigger "spread"
of the bitrate than 2 pass vbr. I've not tried it myself.

Are you using a bitrate analyzer to the the results?

[Kika]

TMPGenc offers a wide range of VBR-Modes. 3 for single Pass and two for 2Pass-Encoding. And they all are very different.

2Pass and CQ are using the same mechanism - they are trying to give all the Frame-Types a similar Pcture-Quality. With the Q-Setting you are telling the encoder, how to spread the Bitrate.

2pass (old) and CQ-VBR are using an other way. They do prefer the I-Frames over the other Picture-Types. Sometimes, this will be better, sometimes not. That depants on the Source and on the Bitratesettings.

MVBR is a very special Mode, too hard to descripe it without writing half of a Book...

OK, back to CQ: This can be used very different ways: One is to set Standard-Bitrates and vary the Q-Settings. That's what DVD2SVCD is doing. An other way (and the Way i use) is to set Maximum Bitrate much higher and using lower Q-Settings. But this is only suitable for encoding to DVD or Non-Standard SVCD, because you can have peaks far away from Standard-Bitrates by using this methode. The advantage of this way to use CQ is, that it makes much better results possible, but it's very hard to find a setting to fit a defined Files-Size.

Hope that helps a little.

[Brother Darrell]

...An other way (and the Way i use) is to set Maximum Bitrate much higher and using lower Q-Settings. But this is only suitable for encoding to DVD or Non-Standard SVCD, because you can have peaks far away from Standard-Bitrates by using this methode. The advantage of this way to use CQ is, that it makes much better results possible, but it's very hard to find a setting to fit a defined Files-Size.

I have read that you can somewhat guess at average by using a simple formula. I have tried it only once but it came pretty close to my expected average bitrate...mind you, this is using CQ ONLY, may work for AUTO VBR too...

If you have a maximum bitrate of say 2600(my SVCD encodes) and you want an average bitrate of 2000, first you use a handy dandy bitrate calculator & find what the expected size would be for an average of 2000,then,what the size would be for 2600. With the quality bar at 100, start sliding it backwards 4 steps for every 100meg you need to reduce it. I assume 1 step may equal 25meg either way.
Like I said, I only tried this 1 time and it worked.

However, going back to something you said about MVBR, This may be my only way to tweak the bitrate. I have never used it but it's worth a try.

Thanks.

[Brother Darrell]

Originally posted by davexnet
I've heard it mentioned that CQ does provide a bigger "spread"
of the bitrate than 2 pass vbr. I've not tried it myself.

Are you using a bitrate analyzer to the the results?
No, I'm just looking at the log file after the encode is done.
It's all right there, GOP by GOP.

[Kika]

MVBR isn't easy to descripe. It is the Encoding-Methode for professional Users. Because you can change anything WHILE encoding: Min- nad Max-Bitrates, GOP-Size, Q-Level, Matrices and other stuff.
MVBR is the reason why i asked for a proper analysing Methode to find High ond low action scenes in a Videos a couple of weeks ago. Without such an analysing program (maybe AVISynth can do that, but i'm not a programer) there is no way to use it meaningful on longer Videos.

[r6d2]

I understand that the CQ parameter in TMPGEnc CQ mode behaves similar to Bias on CCE. CQ=100 means "don't gamble, keep the BR at max". Lower CQ means "feel free to experiment". The lower the CQ, the freer the experiment. Too low CQs behave rather erratically though, in terms of BR.

CQ in a sense does not "spread better" the available BR, because it does not know what it is (it's only 1-pass). But it does a good job at keeping the quantization somewhat constant, generally at a higher level than VBR 2 pass, but more stable.

I have found that setting the BR to max by spec and then finding a CQ with some sort of prediction or iteration, gives good results. It is advisable, however, to keep the CQ >= 70. Otherwise, use more CDs (you can always lower the max too).

Just my 2 cents.