Log in

View Full Version : Problem with xvid movement compared with divx (i include samples)


Alxemi
3rd February 2003, 20:27
Hi all
I´m newbie in the xvid world, but i´m so exciting of introducing it in my OGM rips :)
You can download the samples while you read ;)

http://alxemi.cjb.net//samplesd9

The problem is that my first xvid gives me a strange problem... if you have some time and good eyes maybe can help me ;)
First of all, my codec configuration:

Using latest Kopei build (26-12-2003)

1 pass:
Motion Search Precision: Ultra High
Quantization type: MPEG
Four CC: Xvid
MAximun I-frame: 300
Minimun I.frame: 10
Chroma motion enabled
Quartelpel enabled
GMC enabled
B-Frames 3/150/100
Packed Bitstream enabled
DX5.0 b-vop compatibility enabled
Other options Default

2 Pass int:
Size: 631493 (+-1140)
Quantization type: New Modulated HQ
Quantizatizers:
Min I-frame: 2
Max I-frame: 6
Min P-frame: 2
Max P-frame: 16
I-frame bost 0%
Discard first pass enabled
Below I-frame distance: 10
I-frame bitrate reduction: 30
Curve Compresion: 0/0/240
Payback Proportionaly enabled
Other options default


I think this is a good configuration, ¿what do you think? ¿Any big mistake?
Later i´ve readed that b-frames are not a good idea for animated movies.. i will take count in my next animated rip


The video codec very slow (6-9 fps) compared with my normal divx speed (9-11)

Then, when i see the movie, i see micro-stops. The movement is not fluid, is like it stops a very-very little time to continue playing and microstops again...
I think the better sample is the dogs sample. You can see how the scroll is not fluid better that with the cat-head-moving

Then i codec the hole movie again with divx5.03 and this settings:

Multipass 1st pass:
1140 kbps
max keyframe 250
Psychovisual light
Quarter Pixel
GMC
BFrames

Multipass nth pass (only 2 passes)
Bitrate modulation 0

And the movement is perfect. Something is going wrong with the xvid version, but maybe is the coding, or the playback, or my system... i dont know, but i´m sure that, in my system, the difference exist. With random playbacks i can cleary difference the two versions.

I have a AMD-Duron 800 with 128MB of SDRAM, voodoo 3 3000 16mb running Windows98SE for Rips with DirectX 8.1

For playback i use bsplayer, last build 494 (in nightly builds) with last alpha ffdshow filters (20030103)

Thank you all for your time and dedication in this forum. It is with no doubt the BEST DVD-backup related forum in the net.

Teegedeck
3rd February 2003, 23:34
I'd say 'packed bitstream' is the culprit.

Alxemi
3rd February 2003, 23:48
Thank you very much for your reply, Teegedek!
I´ve seen the documents i used for my rip, and i don´t understand why i checked this box on... it is off in all my docs.

So did you seen the samples and notice the micro-delays? I began to think that the problem was my system, because everybody tells me the samples was right...

Teegedeck
4th February 2003, 07:37
Nope, the samples really play well for me.:)

It's only that your description reminded me of something that I had seen long ago when trying to playback packed bitstreams with ffdshow. Do you use it by any chance?

sysKin
4th February 2003, 10:22
Originally posted by Alxemi
I think this is a good configuration, what do you think? Any big mistake?Well this is not a big mistake, but still - and I'm pretty sure that many people don't know that. Any kind of modulated quant type will not work with b-frames (with 'max' setting different than -1).
As modulated quantizer is illegal in mpeg4 standard (even if works with known decoders), new encoding functions - the ones for b-frames - don't support it.

If you still attempt to use modulated, strange things might happen. I'm not sure, but if I remember correctly - quant type will only change at intra frame. Some scenes (not frames) will use mpeg, some will use h263... and the choice will be quite random. Anyway, this result is not what you need.

Best regards,
Radek

Alxemi
4th February 2003, 23:26
Well i have now code a short THX intro without the Packed Bitstream option cheked... and it Works!! No more micro-stops.

I think i just readed the description of the option and i thought.. humm that sounds god :)

About the quantization... I had no idea about the incompatibility between bframes and modulated HQ... I will search the forum.

Thank you for your replys!!