Re: [Xvidcap-users] About, capturing the framebuffer...
Status: Beta
Brought to you by:
charly4711
From: danalien <dan...@sl...> - 2005-05-18 16:41:53
|
On Wednesdayen den 18 May 2005 11.08, Arne Caspari wrote: > Hi all, >=20 > Such a feature will ( almost certainly ) not go into unicap simply=20 > because it is not in the scope of such a library. Well, I hand to ask, anyway.=20 as it would have been nice if Karl had some lib with good API's to make=20 xvidcap r0ck :-) But, even though, maybe someone else might add support via the plugin-syste= m=20 mentioned in the summary of unicap (on sf.net...) for 'static devices' too = :-) =20 > Having said that I would guess that simply capturing the framebuffer=20 > would be rather simple - this is what all the screenshot application do=20 > AFAIK. But it is a very slow way to capture the graphics since the=20 > framebuffer must be accessed via AGP which is too slow. oh, I thought they hooked up to X ... Q; is writing to the framebuffer (FBDev) as fast as reading from it? =20 > The fast capture applications for windows hook up a mirror driver which=20 > mirrors all gfx requests such that the driver can capture the graphics=20 > without the need to read the whole frame buffer. a 'mirror driver'? ... can that be done with non-f/oss drivers, eg, nvidia? /* thinking only in linux here.. */ > Doing something similar on X you would need to use a mirror X server.=20 > While there are some, got some more info about how? to read? ... googling didn't quite give the s= ame=20 interpretation of the word 'mirror' :-) =20 > I do not know if any of them provides the required functionality.=20 > At least this is a non-trivial task.=20 Karl H. Beckers wrote: > Hi there Dan, >=20 > for starters, I have no idea how the fraps people are doing their=20 > capturing. I'd very much assume that they hooking up to the DirectX=20 > stuff which on Linux would mean messing around with X11 or prolly=20 > libmesa or somesuch. oooh oohh, I know, I know... they hook up some tiny, tiny, tiny mirrors to= =20 their hardware, and copy/dumy *that stream ... =3D) j/k > Though I would love to have such a feature, I don't think I'll ever=20 > start implementing that myself (because I'd never get to finish it ...=20 > as anybody who's following the progress of this project can confirm in=20 > light of the slim time I get to spend on it :( ) unless it comes wrapped= =20 > in a handy API ... yeah, as someone who has followed the project, one got the general hint you might be a tad busy.. > Unicap might be one such API. Have you any smart ideas about this, yet? As it turns out, it isn't one such lib/API :/. though, one hope, might still exist if it's possible to craft a plugin usin= g for unicap's plugin-system, to support gfx-cards as 'capture devs...' :) =20 =2E.. but then again, one has to weigh in if capturing the frambuffer vs. t= he 'mirror driver' is worth the hassle... =20 =2D-=20 // with regards // UID :: danalien :: <${gofigurewhatmyemailis}@slicker.org> // KID :: 0x446AA288 =2D - - /\/\ (^.^) (")") *This is the cute kitty virus, please copy this into your sig so it can sp= read |