back into the fold....whats changed?

[NSanity]

the last time i used xvid was about 8 months ago... Although i have been loosely following its development i'd like to know whats happened...

In theory, what should give the highest quality (even if its the slowest encode).

I tried XviD-04102002-1-STABLE the other day... and it crashed out on second pass (New Modulated HQ, VQ=6 bitrate was 850ish)

Is H.263 still the best or has something surpassed it for the maximum compressibility/quality settings?

Thanks in advance guys...

[Doom9]

don't use modulated quantizers anymore... they are not fully mpeg4 compliant. And for safety use VHQ1 only.

[mf]

Originally posted by Doom9
don't use modulated quantizers anymore... they are not fully mpeg4 compliant. And for safety use VHQ1 only.
For B-frames that is true yes. Without b-frames modulated quant is entirely safe and compliant. And I agree about VHQ1.

[PowerMacG4]

@mcf

I think he is correct.

[kaitsuburi]

@NSanity-

Iago's and Drebel's guides in the sticky give a good idea about some of the new features and how to use them. If you haven't seen it yet, here's a recent thread about alternative (custom) matrices:

http://forum.doom9.org/showthread.php?s=&threadid=50673

Of course, make sure you do some visual testing to convince yourself.

Defaults are great. vhq=1 and chroma me are also generally recommended. Then you can do a couple of different things depending on your source/compressibility/what you want as output.

-kaitsuburi

[sysKin]

Originally posted by mf
For B-frames that is true yes. Without b-frames modulated quant is entirely safe and compliant. And I agree about VHQ1. Oh, mf-baka, you're wrong. Lemme tell that one more time:
- without bframes, modulated quant is not mpeg4 compiliant (it's not allowed to change quant type during a sequence)
-with bframes, modulated is still not compiliant (for the reason above) and doesn't work correctly: xvid will only change quant type on intra frames (which is still not allowed by the standard), while the idea of modulated quant requires to change quant type more often (possibly every pframe). As the result, some _scenes_ will get mpeg quant type, some _scenes_ will get h264 quant type, and this is will not bring the effect of improved quality. It will still work, sure, but the quant distribution will be mostly random.

[Teegedeck]

What's changed? Hm, since I couldn't use XviD for months, really (due to little time and insufficient hardware) I was in a similar situation and think XviD's present status sums up like this: It's gotten so much better! :D

Really. B-frames work nicely in scenes without motion (thanks to sysKin, I believe), now, and the look of quarterpel is superb if the bitrate is reasonable. It's so very good that I find nothing sensational anymore in WM9's claim to deliver DVD-quality at 1 MBpS. You may know how much I'm inclined to encode everything at full resolution...

So, I'd say one can finally keep the results if one experiments with B-frames, quarterpel and VHQ (VHQ=4 is very OK, also. It was before, because it saves the most bits in a 2-pass-encode, and also it's been revamped again so that none of the VHQ-modes gives a PSNR-decrease, anymore. So I've read.) And I'd use MPEG-quant for low bitrates, also.

Just one thing: Don't use lumi-masking. The old one actually saves bits, but it's not necessary because the visual cost is too high and XviD is very efficient without it, already. The new one (RefDivX's code) simply decreases PSNR, i.e. less crispness at high bitrate, combined with larger frames, and it doesn't help anything at low bitrates, either. Sorry to say that, really, really sorry.

[Didée]

Originally posted by sysKin
... some _scenes_ will get mpeg quant type, some _scenes_ will get h264 quant type...
Really? WOW!

;)

[mf]

Originally posted by sysKin
Oh, mf-baka, you're wrong.
You're absolutely right :o. Not-working and non-compliant is so easily mixed up :D. Sorry guys! Always believe sysKin :p.