[XviD-devel] just a question

[Marc FD]

> There is no official "ritual" of how to become a core coder and of course
> we can always use more people. If you have good ideas and also write good
> and portable code for these, you can always help.

cool.

> I don't know how much you know about the internals of
>MPEG-4 encoding

nothing

> or XVID itself.

everything ^^

i know basically how MPEG-4 works, and i understood amost all
MPEG-4 tools implemented in XviD by reading the code ^^
i recently played with MPEG2dec's internals too.

> Image processing is only a small part of this, but of
> course it's an important part.

yes, and i'm not limited to image processing at all !!!
i'd like to do the fun stuff, help implement new features,
make XviD not only the best MPEG-4 encoder out (it is already)
but so good that DivX 6 will be outdate even before they release it !!

> Currently, many things are new in XVID, and there are many ideas for
> improving existing features as well as including new ones, and there are
> surely not enough people to work on everything at the same time.

pete talked about reduced vops, and i dunno why, but i'm attracted by this.
it'ld be great you could explain me in details (or point me to the info)
about
this part of the MPEG-4 spec.

>others are more "image oriented" (like prefiltering).

that's a job for me ^^
i'll explain you my ideas about that in a new email-"thread"

> Some are completely new (like dynamic B-frame decision, I
> don't know any other encoder using it).

that's one of the coolest parts of XviD. a genius idea from syskin.
i've cool ideas to extend it to do smart preprocessing ^^
it's exactly like the 2pass ME that i wanted ^^

> There are also many things to improve beyond the core itself, like
> distributed two-pass which we were discussing recently.

that's funny, i exposed exactly the same idea for users on a french
forum who wanted to do SMP/cluster encoding a month ago.
if you can read french i can give you a link ^^
BTW, did you think of using margins of <max-pframes> length
in 1st pass cluster encode to resynch at scene changes for 2pass
scene based distribution ?
using this, you can easily encode a 2h movie on 10 computers
(or SMP systems)

movie splitting/stats merger/spliter and movie merger can be done
using opensource code. if you would need to code it, i would not use
avi, but directly ogm ^^. (because it's cool and opensource)

> So, just tell us what good ideas you have in mind! Maybe somebody looked
> as something similar earlier and you can discussion and even start
> implementation. Many good ideas only came up because of discussions, e.g.
> here on the list...

i'd like to implement advanced MPEG-4 features, but i don't have the specs.
BTW, it seems to me that skal lives very near to my location (orsay,france)
so if he want he could teach me a lot of things ^^

i think prefiltering will be cool in XviD, i'll expose my view about this in
a dedicated mail.

i'm joung, i've always tons of ideas, if i knowed more the spec,
i could find even more cool way of using it.

>i consider anyone who is active on xvid-devel to be a ~core developer~,
>except for people inquiring about 'xvid audio'. seriously, if you feel
>like contributing you do not need to ask for our permission or
>acceptance.

it's just because i want to code a _lot_, so i prefer be sure it would be
integrated ^^

>my question: is Marc FD the same chap as 'mf'?

it seems 'mfa' and marco (AI) is the same chap,
my name's marc fauconneau, so 'mf' is possible too.
i'm a bit lost here ^^

pete, can you say more about reduced vops ? i'd like to know more
about it. (it may be something i'm seeking after in the spec)

thx all ^^