Quality difference between 1.0.3 and 1.1?

[compunerd632]

I just started ripping my dvd collection so I can have all my movies on a hard drive for easy access by the Xbox. I'm using xvid 1.0.3 right now, but since the 1.1 beta came out I was wondering if I should use that. I've only got 5 movies done so far, so it's not a big deal to redo them. Basically I'm asking if there is a noticable quality difference between 1.1 and 1.0.3?

[SiXXGuNNZ]

Originally posted by compunerd632
I just started ripping my dvd collection so I can have all my movies on a hard drive for easy access by the Xbox. I'm using xvid 1.0.3 right now, but since the 1.1 beta came out I was wondering if I should use that. I've only got 5 movies done so far, so it's not a big deal to redo them. Basically I'm asking if there is a noticable quality difference between 1.1 and 1.0.3? how bout encoding them yourself, then posting the results. I would say go with 1.1 beta, but then again, it is all about opinions.

[Sharktooth]

From the experience i had with the pre-beta builds i can say there is an improvment both in speed and in quality.
Ensure you enable VHQ for B-Frames if you want the best quality (it will slow down the encoding speed a bit).

[compunerd632]

I redid one of my movies using the 1.1 beta and used the same settings except I turned on VHQ for b-frames. There didn't look to be any difference while watching the movie, but if I looked frame by frame the 1.1 beta seemed to have less ringing on the edges of some objects. I also looked at an explosion that had some blocks in it. Neither one seemed to be better in this case. Some of the frames were pretty clear while others were blocky and the two different clips weren't always blocky on the same ones, but in general seemed to behave about the same. I did notice a slight gain in encoding speed as well. So I guess I'll keep using the 1.1 beta, but it's not really worth it to go back and reencode your movies. Hope that helps out others.

[Sharktooth]

The quality difference can be spotted on those encodings with lot of b-frames (since the vhq for b-frames affect only ehr... b-frames).
The more the b-frames the more quality difference there will be.
That doesnt mean you should change the default b-frame sens. or max cons. b-frames value. The codec chooses when to place them.
If you know what you're doing however...

[wolfsoft]

In my feeling, the quality of 1.1 is most like wmv9. But the size of file couldn't control well like 1.0.3. It always makes a movie by different size than I set. For example, I set the quantizer as 2.5 in single pass mode,and the quantizer was 2.7 or 2.8 in finally. Because the average quantizer of I and P is 2.5. But 1.1 wouldn't calculate with B frames. So the value would be larger. If I open the VHQ for Bframe, the offset would be more large.

[chilledoutuk]

@wolfsoft

I have not had any problems with the latest build and its rate controll your problem is that your not encoding your videos the way the devs intended.

if you want to hit a certain filesize then use quants 1-31.

@all

i noticed a big improvement with VHQ on bframes when encoding full anamorphic sequences at relatively low bitrates.

I have been getting good results with this beta and have been very happy so far with the quality of encodes made with it.

[wolfsoft]

@chilledoutuk

I always set all of quants min as 1 and max as 31 until I used xvid. That's why I feel strange.

[Sharktooth]

try with integer values or do a 2 pass encoding to hit the final size.

[AsTimeGoesBy]

Originally posted by compunerd632
I just started ripping my dvd collection so I can have all my movies on a hard drive...
I'm doing this for the last half year. Will this ever end...? ;)

Originally posted by compunerd632
Basically I'm asking if there is a noticable quality difference between 1.1 and 1.0.3?
My very personal opinion after comparing some of my test samples: 1.1b is really better :) (although i guess that was intended!?;)).

Generally no need to re-encode a Xvid 1.0x collection, but as far i have experienced Koepi's Xvid 1.1b has made some notable progress on scenes with (very) fast movements. One little impression you may win looking at some 2 pass Finding Nemo frames (http://n.ethz.ch/student/fhoesli/video/dvd.fn/). - I'm sorry, there are also frames encoded by another codec that will attract maybe your attention...
However, fast and violent motion scenes are one thing. If your preferences are on stable images on steady or lower motion scences i wouldn't leave the Xvid trace.


-----------------
Edit:
Please note, the hole film was encoded at 2250kbps resp. 2.5Mbit/s - this scenes and these images not!
This dark scene (where the imagea are taken of) had only an average bitrate of about 800kbps!
So within that 2 pass encoding >2MBit/s Xvid (specially v1.03,but indeed also DivX) are spending to few bits on that dark fast motion scene while other scenes in the film are almost 'unfair' glass clear.
Soon I will put the AVI files too to leave a better impression about the benefit of Xvid 1.1b in that scene.

[Sharktooth]

ATGB: you should do something wrong at least with xvid... i re-encoded the clips and the image quality is much-much-much better than yours...

[AsTimeGoesBy]

need a minute to celar this...
my head is smoking...

[Sharktooth]

ok, gimme some time (i trashed the files) and i will post the encodes at 2250kbits.

[AsTimeGoesBy]

I had really had to add an important note...
However i guess it gets evident that Xvid v1.1b is better than 1.03 even at that difficult scene.
About stability and encoding time i can't say much, but personally i have begun to use 1.1b after my own litte comparison just because of the quality reasons. If i get some more server space i will post a link...

Sharktooth, thanks for your correcting input and sorry for all that inconvenience.

[Sharktooth]

settings: default settings except I-frame int. = 250 + Chroma opt + Adaptive quant + VHQ for b-frames + quant matrix = EQM V2 + PAR 16:9

clip: http://ebola.gamersrevolt.it/temp/clip_09sec.avi

[AsTimeGoesBy]

Thanks for the file! - I'm looking forward to to watch that cross-check this evening. - A costum matrix was used, that will be interessting...! :)

Adaptive Quantization is used too!? - In an (older) Xvid guide i have read that should only be used only for lower bitrate (<1MBit/s) encodings. Is that advice that obsolate since v1.1b is available?
I mean Adaptive Quantization will stress dark scenes once more...!?

[Sharktooth]

I enabled it just coz you had it enabled in your test... AQ is more effective at low bitrates but in certain encodes it could be left enabled without problems (this is not the case though).
Btw, yes, it will stress dark scenes.

[AsTimeGoesBy]

Adaptive Quantization should be something like a 'psychovisual effect' if i remember right.
So far i have nothing against this as long i can't regonize the bitrate savings.

By the way, the cusom matrix you have used on the sample has a nice effect, something like 'film effect' realizable by DivX configuration.


To get back on the topic i have posted some Xvid samples (http://people.ee.ethz.ch/~fhoesli/video/test/?skip=divx) for all who distrust Xvid 1.1b or just want to get an impression.
(sorry for the schoolish touch of the page, but indeed it's intended for this, to passively bribe my multimedia-fool teacher;))

[Sharktooth]

If you like the CQM you can download it from my signature (EQM V2).
BTW we have some problem with the DNS server so i had to shut id down until i get it fixed (that means tomorrow).
Psy enanchements usually lower the details on very dark (where the details are "hidden in the shadows"...) or very bright areas to spare bits to place where details are more visible. So, if the scene is completely dark, AQ will not perform so good...