[XviD-devel] [QUESTION] how 2pass2 plugin work?

suxen_drol suxen_drol at hotmail.com
Wed Apr 9 01:15:09 CEST 2003


hi,

ive commited further changes to 2pass1/2.

original and scaled frame-lengths are now stored in the one stats file.
when the 2pass2 plugin is initialised, only one stats file is specified.
if the target bitrate (bits-per-second) is non-zero, the xvidcore will
scale the curve, if not, the scaled_length field the stats file will be
used.


2pass1/2 todo list:
* futher testing;
* error checking in 2pass2. theres many places where things can wrong.
* credits support (or something better...?)
* update the bloody vfw
* ratecontrol chaining. e.g. "divx(tm) multipass(tm) encoding"
the ability to perform to write out 1st pass stats, whilst performing
the 2nd pass. ie.
    xvid_enc clip.pgm -pass1 log.txt
    xvid_enc clip.pgm -pass2 log.txt 800000 -pass1 log2.txt
    xvid_enc clip.pgm -pass2 log2.txt 800000 -pass1 log3.txt
    xvid_enc clip.pgm -pass2 log3.txt 800000


i think dev-api-4 is shaping up to be merged into CVS_HEAD soon. the
adaptive quantization (qscale table) api stuff needs modification, but
otherwise is almost good to go.

On Mon, 31 Mar 2003 15:26:28 +0200 Dirk Knop  <dknop at gwdg.de> wrote:
> Isn't statsreader (for the windows side only yet) exactly doing that?

yes and no. the nandub stats fileformat has been ditched, in favor of a
plain text file format.

> That would be fun, that way we'd "know" from the codec side that this file is
> already scaled or better has a scaled curve added(we should increase statsfile
> version to -21 then or something like that).
> 
> So we only select 2pass, and if we scaled the curve the codec knows that we're
> doing a 2nd pass.
>
> Or is that BS?

nope!

-- pete




More information about the XviD-devel mailing list