Pink stuff using DivX 5

[meta]

Hi all,

Im using the PX3 Mpeg2avi for encoding video.

With 4.12 codec this is sweet. But for 5.0 I'm getting
these wierd pink shapes predominantly around the bottom
right corner of the avi.

Sometimes this pink "ghost" seems to pixillate (cant spell that)
and moves as objects in the avi move.

Anybody know how to solve this or where to start?

[techno]

I personally do not like DIVX v4.xx and v 5.xx.

I have you used a crack for this? if so this is the problem. I do not know what else to say

Techno

[khp]

What resulution are you using ?, both width and height must be divisible by 16, otherwise you get pink stuff at the edges.

[electrongunner]

I get pink stuff too...could be the movie dimensions...

I'm a seasoned ripper that's ripped movies every which way and I was initially pretty excited by the quality I got ripping with DivX 5.0 in 2-pass mode using XMPEG 4.31. XMPEG is a major modification of Flask and really looks and works great and seems pretty damn fast. However, after finishing the AVI, I scanned through it and discovered the pink artifacts you mention. It may be due to the fact that instead of ripping to a final size of 640x280 (cropped), I ripped it to 638x280 because the movie had a couple of columns of black along one side, so I decided to crop them out rather than waste space encoding them. I have a feeling the artifacts may have come from using a width of 638.

I don't believe the message about needing to have dimensions divisible by 16....I think it just needs to be divisible by 8. I've always used 640x280 for widescreen movies and never had a probelm after ripping close to 50 movies even though 280 isn't divisible by 16. Since both 640 and 280 divide evenly by 8, but 638x280 doesn't, I suspect that's where my artifacts came from. I'll try to remember to post a followup to let you know, but I'm tempted to switch back to DivX4.11 just to ensure I avoid another failed session. Maybe if I had a Pentium 4........sigh.

[meta]

Pink Stuff

Thanks to those guys above the prob is solved and yes it was the dimensions ivisible by 16 works fine - no pink stuff.

Initially i though well back to 4.12 however since this revelation I think I will tinker a little with 5.0 and see what happens.

They claim that 5 is faster to encode with and more compressable than 4. After using a 2 pass for both a number of times it seems to me that that claim is bullshit (pardon the Australian).

As for image quality my eyes seem to tell me that perhaps just perhaps the image is sharper holding constant as many variables between the 2 codecs as poss...again time will tell and Im not getting into that debate its been done before.

Anyway thanks for the posts guys and happy encoding!

[khp]

They claim that 5 is faster to encode with and more compressable than 4. After using a 2 pass for both a number of times it seems to me that
that claim is bullshit (pardon the Australian).

When you turn on one or more of the three mpeg-4 options (QP,GMC,B-Frames), both encodeing and decodeing will slow down significantly (Quality at the cost of speed). If you turn off these options (and ofcause all the other preprocessing options) encoding should be a bit faster than divx4.12, but that won't give you any quality improvements.

So in short divx5 is both faster and better quality than divx4.12, just not both at the same time

[meta]

Divx 5 speed and compression

Fair Enough. Have you actually tested this yourself?

How much of a difference do you think QP,GMC,B-Frames makes?

[khp]

Yes I have done a few tests, judgeing quality is rather difficult, so I usually use 1-pass Quality based encoding and use filesize as a preformance measure (assuming that Quality based encoding always produce the same quality)

I think this thread is one of the better ones on this subject. But there are several related threads in the same forum.

I am currently working on a tool that will try to measure the the video quality of a avi file, I will release the source code when I have a minimal working solution (Hopefully in a week or two).

*edit* Please note that Q-Pel has been reported to produce visible artifacts (moveing textures), this was not known when I posted my first results.

[angel]

I have the same prb with getting this pink image, and I'm encoding on SBC divx. Is it because the screen size is not divisible by 8 also? But the original file has screen resolution of 640 X 360, and 360 is not divisible by 8.

I've been encoding widescreen movies using 560 X 315 resolution, sumtimes the pink stuff comes out, and sometimes it doesn't (really confused now)

[khp]

Originally posted by angel
But the original file has screen resolution of 640 X 360, and 360 is not divisible by 8.

That's news to me. I've got a degree in math, so I really should know about this.

Originally posted by angel

I've been encoding widescreen movies using 560 X 315 resolution, sumtimes the pink stuff comes out, and sometimes it doesn't (really confused now)

Use 560*312. Or even better 560*320, which makes both height and width divisible by 16, which is optimal.

[angel]

Thanx, I used the resolution which are divisible by 16, hopefully the prb is gone. Well, I just browsed thru the video quickly, so far didn't see any of those pink artifacts...