can somebody explain this 2 me?

[yangus]

Ok here's the scoop: I have 2 divx movies The Matrix and Blade, both of which were downloaded off the net. I watched and examined both of them and they both look absolutely amazing, only with careful examination do i see little artifacts. However what really amazed me were their small file size, blade = 625 MB and Matrix = 578 MB. and each movie were about 2 hours long. i ran the calculation in a bit rate calc and supposingly the bitrate would have been 400~500 kbps. however that's impossible, b/c the quality is just absolutely amazing, even in full screen. and the audio quality was awesome also. btw i noticed they were encoded in divx 3.11 alpha, which is even more amazing since divx4 is suppose to look better. can anyone explain this to me? i just encoded rush hour 2, 1 1/2 hour in divx 4 and the quality is comparable to the matrix rip, but the movie is shorter! how do u expalin this?

[khp]

First of all these movies were most likely done with SBC encoding in nandub, which will give u the same (or even better) quality as divx4 with divx3. Next they might have been done with vbr audio at 112 or 96kbs, which in my calulations should give u a video birate of 550-650.
Futhermore they might have used alot of noisereduction and smoother filters (noise really kills compressability).

[glenn]

Those movies were likely encoded with Nandub, which is a tool that gives comparable quality to divx 4 through the use of divx 3.11. However, divx 3 is quickly becoming something of a dead end; there is no further development. The people that once made divx 3 what it was are now working on divx 4, which will only get better. At the moment the two solutions give about the same quality, tomorrow it might be different.

However, perfect quality for a 578mb Matrix rip? Not likely. Try turning off your codec's post-processing filters, and see what happens...

[yangus]

yea i kinda figured they probably used nadub after reading the "divx4 sux" thread. however i checked the settings for the mp3 and it seems to be a constant 128kbp and even with the postprocessing turned all the way down it still looked amazing. i can even post screen shots if u guys don't believe me :D so if 3.11 movies using nandub can surpass 4.12 movies, can't u use nandub with divx4 to produce amazing movies? nandub does support divx4 right?

[The FX Files]

No.

[JayTray]

No....... you must Use Virtual Dub for DivX 4.x. Nan Dub only works w/ DivX 3.11.

Yes, Nan Dub does give excellent results. There are people who think DivX 4 (2-pass) works better than Nan Dub and DivX 3.11. I have done both and have found DivX 4.x is better quality. However, DivX 3.11 (using Nan Dubs SBC) is much better at calculating the Bit rate than Virtual Dub is with DivX 4.x. Even using gordon Knot has not solved the file size issue. Anyone Using Nan or Virtual Dub to encode most likely uses Gordon Knot(Gknot for short).

If you haven't you are working to hard. Gordon Knot takes all the hassle of encoding. To read a complete guide on using Gordon Knot and Virtual Dub (DivX 4.x) Click Here! (http://www.doom9.org/divx4-avs-main.htm) For a complete guide on using Gordon Knot and Nan Dub (DivX 3.11) Click Here! (http://www.doom9.org/gknot-main2.htm) If you follow this guide you will have a perfect DVD Rip!

Good Luck

[ChristianHJW]

... there are in fact a few real good rips on the net .... the key to do them is not using Nandub/3.11 or Vdub/DivX4 2 Pass but filtering, Baby, filtering ...... and there are professional video filters existing that will do fantastic things in terms of improving compressability .... but this can clearly be seen as it will remove a lot of picture details also ..... but you dont get macroblocks even at real low bitrates.

The best compromise i found is to use 'light' filtering in GKnot ( tempsmooth 2,1 ) with 'insert filter before resizing' ticked. Its slow ( SMP PIII 800 will do this for me ) but gives good results ...

[yangus]

oh man that's really 2 bad u can't use nandub with divx4. i wonder what better result one can get with those 2 combined 2gether:p and btw JayTray, i never have a issue with filesize when it comes to encoding in vdub. my movies always come out to be a couple of mb off of what the bitrate calc says. just saying...
and ChristianHJW, do those wonder filters u talk about come along with gnot? i would like to give them a try and c!

[JayTray]

2~Yangus

Do you use two pass or one pass. Are you talking about the DivX 3and4 calculator.

I wonder....... If I used that to calculate the bitrate and then use Gknot to create an avs. Using the bit rate recommended by the calculator rather than Gknot. I wonder if the file size would be more accurate.

I'll try that.....

[ChristianHJW]

Originally posted by yangus
and ChristianHJW, do those wonder filters u talk about come along with gnot? i would like to give them a try and c!

No, i am talking about professional video equipment, DSP based, like being used in Studios ..... to get an idea of what i am talking about, you can do the following in GKnot :

When your .d2v project is loaded and you're going to save the AVS script file, tick

'light filtering' and
'insert filter before resizing' .

Dont have GKnot installed here on this OS, so wording may be a bit different.

Then edit the AVS script file from 'tempsmoother (2,1)' to 'tempsmoother (3)' and save it. Make sure this line is coming before the 'bicubicresize (xx)' line, so its in the right order.

You'll need plenty of CPU power or plenty of time now, but you'll notice a certain improvement, especially when doing the following tests :

1. Encode a certain part of the movie in '1 Pass' mode, with mQ = MQ = 5 ( constant quantizer encoding ) f.e., bitrate 6000 kbps ( to disable rate control settings ) . Make 2 versions with same bitrate and settings, but one with filtering and 2nd without. Report file size here.

2. Make a normal 2 pass encoding, but with wrong resolution settings like full DVD res, resulting in 'bits/pixel*frame' values < 0.16, sharp bicubic resizing. Of course, normally you would never do so but we want to drive the codec to its extremes. Make sure you do a longer encoding ( min. 20 minutes ) and choose a part of the movie where both action and low motion are present, to make 2 Pass sensible.
Again 2 versions, with and without filtering. File size will be ( roughly ) the same now as dictated by bitrate settings, but you'll see big differences in terms of macroblocks, etc , i promise ...

[everwicked]

Size predictability with divx 4 was solved with divx4log. At least it works for me.

Check my guide for details. www.med.uoc.gr/~diamantop/ (http://www.med.uoc.gr/~diamantop/)

[yangus]

here's a pic of the matrix, which unfortuantly i had take in vdub with snagit, so the quality is poorer than the actual.which brings up another question, whenever i use a pic taking program (ie hypersnap) to take pics of divx movies, it always give me a black image instead of the actual screenshot. can anyone explain this?
2 JayTray, i use 2 pass divx4 and the bitrate calc that comes with gnot. works fine every time.
and ChristianHJW, i'll give this a try when i actually get the chance to, right now i have too many movies to encode before school start!

[yangus]

ok what the heck, the image did not attach! well just image the dvd version of the matrix where neo was showing morphious some "kung fu" moves and he's doing that backward flip. tone that image down just a little and there's the pic. :D

[JayTray]

~~~~~Yangus....... Try reattaching those images. Maybe they were to big. How big were they?

I think I figured out what i was doing wrong with GKnot. You know how AC3 (6ch) is way bigger than MP3 file in MB. I would Encode the Audio Seperate and mux them when I was done. Because more bits a freed up for the video, I would just add a file size. But I guess Gknot calculates the audio in there some how too. So when it is encoding the final pass it expects the audio to be included in the mux. Or at least that is what I am guessing. Do people encode with "no audio" and have no problems. Maybe it's just me.

Well, Thanks for the advice Christian. I will definitely give that method a try. I wish I could see some samples to compare.
Also, I was wondering what changing the tempsmoother (2,1)' to 'tempsmoother (3) does. I tried the DivX 4 log editor and it helped. I read the guide and it explains it pretty well. I will see if the quality is better with one pass, tempsmoother @(3), than using the regular 2nd pass (no filters) method.
~Hey I know, what if I edit the avs tempsmoother to (3) instead of (2,1), plus edit the Log file and run a 2 pass encode. I wonder what kind of results I would get. I guess I will have to find out!

[serbersan]

Hey, what's the difference between:

tempsmoother(x,y) and tempsmoother(x).

I mean, what's the diference between:

ts(2,1) - ts(3) - ts(2) or ts(1)

Thanks.