Question about OPV

[nashcity]

I use CCE OPV for most of my encodes (if the Q-value is low enough that is) and I was wondering about something. I came across on old thread (Dec '04) discussing the idea of adding a 2nd VBR pass to the OPV method to help with sizing accuracy - I think it was 'tylo' who was talking about it.

Could the bitrate calculated by the OPV analysis be used for a second VBR pass? Wouldn't this theoretically work?

I was thinking that applying an extra VBR pass to an undersized output from the OPV would be quite useless, but applying a VBR pass to an oversized output could result in better picture quality than multipass alone (if you believe in OPV of course - like I do, or more accurately - like my eyes do).

Maybe a method where a determined Q-value from the OPV analysis could be adjusted to get a slight oversize, maybe by subtracting 1 or 2 values from the determined Q-value - then applying the extra VBR pass to size the output correctly. If the bitrate from the OPV analysis can't work, maybe another analysis could be done.

Anyway, I apologize if what I'm talking about is nonsense, I just thought this could be a cool idea of using both OPV and multipass together. I would love to have the option of doing an OPV oversized output followed by a VBR pass for sizing. I personally think the output would look very good in most cases. Of course, I could do this right now, it would just take a few extra steps.

[jdobbs]

If you do a second pass, the calculated Q value and distribution from the first pass would no longer have any significant effect. The second pass would simply take the VAF from the first (OPV) pass and use those values to do its calculations for a standard second pass. The only difference might be if the average bitrate makes a very large change from what was used in the OPV pass -- and that would more-than-likely be a negative.

I think I remember the old thread you're discussing. I think the idea was that you could do an OPV pass -- but then if it came out too large or too small, the second pass could bring it back where it belongs. Which is more related to sizing than quality.

One might be able to debate whether an OPV pass followed by a second pass would result in better quality than a standard two pass... but I've done quite a bit of testing -- and they look the same to me.

[Venom_IL]

Yep, in D2SRoBa it's called a "sizing pass"

nashcity - I actually just asked about the same thing (actually a generalization of it), check the thread named 'CCEront'

Also IMO even if an OPV pass followed by a second pass results in the same quality as a standard two pass, it's still a good option to have as the sizing pass is only a contingency for the now relatively rare case of bad size prediction

Let me put it like this:
You try OPV.
Good (and now common) Scenario - sizing went well, which means you've got great quality in one pass time
Bad Scenario - sizing didn't go well, you do another pass and get lets say the same quality you would've gotten if you had used normal 2-pass mode. But that would have taken roughly the same time anyway (the difference being the RoBa analysis), so in my eyes you really can't lose.

[jdobbs]

Good (and now common) Scenario - sizing went well, which means you've got great quality in one pass time
Bad Scenario - sizing didn't go well, you do another pass and get lets say the same quality you would've gotten if you had used normal 2-pass mode. But that would have taken roughly the same time anyway (the difference being the RoBa analysis), so in my eyes you really can't lose.That was almost the exact discussion of the referenced thread as I recall it. The idea would be for it to work like this:

1. Select OPV for output.
2. At the end of the ENCODE phase a check is made to see if overall size is within some upper and lower boundaries. Normally upper would be that set by TargetSectors and low would be somewhere on the order of 4-5% under TargetSectors.
3. If the size is outside the boundaries, a second pass would be accomplished on all video.

That's still something I'm considering.

[DDogg]

Yep, your memory is correct. A conditional second CCE sizing pass could be used as you already had the vaf, or a conditional nic's recode thing (forgot what it was called) [rejig] could be used if the file was larger. Since the filesize was so close, it didn't hurt quality much and was much faster than the CCE 2nd pass.

Man, that seems like a forever ago.

[jdobbs]

No kidding.

[Trahald]

if its possible.. im a fan of a rejig pass to shrink oversized to compliant .

i use the one pass and whenever its oversized (i have target sectors at default, ie. not listed in the .ini, but sometimes it oversizes.) i just dvdshrink it down to size. its such a small % that video comes out looking the same. yet way faster than second pass. also a second cce pass flattens the vbr curve (at least in my experience) ..

undersized doesnt bother me an iota.

[Sharc]

I would suggest to apply the shrinking in the 2nd pass to the Menu and Extras only and leave the Main Feature intact. That would make the 2nd pass even faster and preserve the quality of the Main Feature. Normally only few MB are missing after the first OPV pass. These can easily be stolen from the Menus and Extras, if these are included in the backup.