Hi, On Wed, Apr 19, 2006 at 11:32:13PM +0400, Evgeniy Stepanov wrote:
On Wednesday 19 April 2006 17:59, Reimar Döffinger wrote:
The stub functions in ass.c might be cleaner to do via defines in ass.h (wouldn't cause useless function calls etc.), not sure about it though.
You mean ass_find_sub, ass_change_track ?
I mean void ass_find_sub(unsigned long long timecode) {} in ass.c vs. #define ass_find_sub(l) in ass.h
2) how difficult would it be to make a vo support this? I think for this it would be preferable if only the ass_draw_bitmap was in the filter and the rendering would be done some place outside, with the ass_draw_bitmap function being called via a VF_ message. Such a drawing function would be easy to implement in the vos as well, and vf_vo could pass such a message on to the vo if none of the filters accepted it. Thus there shouldn't be any need for special cases. Or is there too much data that the other functions currently in vf_ass need and is only available there?
We can render subtitles from somewhere before decode_video() into an alpha buffer, and blend it into image later (in vf_ass or some vo_). But would not blending a large (full video dimensions in the worst case) image be too slow ? What is the possible gain of this ?
You misunderstood. Not a full image, just the same size as ass_draw_bitmap. Or more precisely, the idea is: move everything except the ass_draw_bitmap function out of vf_ass and into libass. There, place all ass_draw_bitmap calls by something along the lines vf->control(vf, VFCTRL_PERIODIC_UPDATE, &struct_with_draw_params); (a helper function similar to af_control_any_rev might be helpful). I guess this is not a requirement, but with the current design I fear it might be very hard to extend. I also admit, that with my proposed changes, we almost get a second vf_osd... Greetings, Reimar Döffinger