3/11/02 New Core for Xvid!

[Dali Lama]

Hi,

The new core for Xvid is out on Umaniac's site. I am going to test it out. There seems to be some type of new 1-pass CBR mode that uses caching? I'll check it out. Hopefully its better.

-h, what do is new in this core?

koepi, nic, are you guys going to produce your own binaries?

bye,

Dali

[-h]

-h, what do is new in this core?

- XviD is now 100% GPL'ed (no Project Mayo code anymore)
- new CBR mode
- new encoding API - we can modify quant matrices, motion search precision, etc. on a frame-by-frame basis
- faster bitstream functions
- patches to motion estimation, search precision 5 should be 20% faster (just for ME - overall encoding speed boost should be ~6%)
- MPEG quantization (both defualt and Custom) both have assembly optimization again

Can't remember much else off the top of my head.

Also, the vfw interface is going to change again.. most likely tonight if the bloody tooltips start working the way the documentation says they should. I don't like Microsoft :)

-h

[Koepi]

-h is still working on the new GUI - which incorporates my changes as well.
So I can sit back and relax (well, not really, the help file and the XOE must be updated).

Anyways, since slp-knot made up some 07032002-3.exe-binary, I guess he'll take over the part of me coding as well, ask him ;)

Still _very_ annoyed (hm, i don't dare writing p***ed o** since it is against the netiquette as well),
Koepi

P.S.: the source still belongs to xvid.org, it wasn't umaniac who invented that ;)

[wing1]

Wait, what happened between 3/4/2002 to 3/11/2002????? xvid development process became quiet all of a sudden.

1-pass CBR??? I AM USING THAT 100% of the time!!!!

[NoLogo]

Hi all.

First, you're doing a great job with the Xvid's stuff.

I just tried the new build (new core), and there seem to be a bug:

I encoded a short bonus (15'), once with an external 2nd pass (using GKnot), once with the internal 2nd pass, and both times i got an avi (the size is correct), but whose very first frame has got a big green stripe, and it makes close my player.
The problem is that i don't think i did something wrong, this is quite easy.
The bonus is an anime one.

Configuration:
1050kbs, 576x320

1st pass:
Motion search precision: 6
Quantization type: H. 263
Max KF: 300 (default)
Min quant: 1
Max quant: 20

2nd pass:
Quantization type: MPEG custom
Other options: see 1st pass

2nd pass internal:
KF boost: 25%
Other options: default

Well, maybe I'm the only one who's got the problem, but maybe this can help.
Sorry i can't tell more about the new core, no way to read the video :)

Regards
NoLogo

[Koepi]

Please try again with first pass using the same custom mpeg quantization matrix.

(well, it's a little bit unlikely, but still possible, that the error comes from there...)

[wing1]

I am quite impressed with the new core :D wow!!!!! However, the xvid.ax filter is not compatible with nic's DS filter and it needs it. Went back to divx4.12 filter and i must said it's very good. Much much faster for capturing.

[NoLogo]

Koepi>

Ok, i tried once againwith the same quantization type (Q.263 for both passes), and i get the same problem. :(
When opening the video with VDub, i can scroll the fisrt frames, and it's quite strange: the very first frame is quite readable (except the gren strip), but after two or three frames, it becomes copletely bugged...
Maybe i did something wrong but i can't see what...
I got two jpegs if you wan to see what it looks like (need your e-mail, or someone's the pictures could help).

Regards
NoLogo