[gst-devel] [theora] Mozilla & gstreamer

[Bill Cholewka]

> > Christopher Blizzard wrote:

> >> 1. An html-based camera + video capture API.  This isn't just an add-
> >> on like interface for JS people but also includes preview + style like
> >> we have with the html5 video element.  Think of it as a file upload on
> >> steroids.  An image interface is easy - just encode to jpg or png or
> >> whatever, and upload it.  Video is more challenging and where the
> >> opportunity is.  I would like to see us include capture for mac,
> >> windows + linux, encode to Theora and upload via the file upload API
> >> that we have today.  This brings video production to browsers (like we
> >> have text + image production today via blogs, etc.) but also continues
> >> the path to make ogg, theora and vorbis part of the underlying
> >> structure of the web.
> >>
> >
> > FWIW, we have a cross-platform capture library, (that we ultimately use
> > with theora), which is at http://sourceforge.net/projects/libvidcap
> >
> > It supports mac, windows and linux.  Presently LGPL, but there are few
> > authors, and re-licensing could be possible if there's interest (we'd
> > have to make sure we're not depending on or derived from LGPL code).
> >
> > Unfortunately, it hasn't been widely publicized or adopted, (and we have
> > some enhancements we haven't sent out to sourceforge lately, mainly
> > because we haven't seen it picked up), but it forms a working library
> > that accomplishes the goals you've set forth.
> 
> This looks pretty neat:
> 
> * ~10,000 lines of code (about 1000 lines of which is colorspace
> conversion, can this be merged with existing code from oggplay?)

We're working on taking advantage of any platform-provided colorspace facilities

> * supports v4l2, quicktime and directshow
> * has timing code for forcing framerate
> * known to work with theora
> 
> Perhaps those updates need to be pushed so people can play with the latest code?

I'll see about pushing anything that's ready. We're also working on loosening-up our overly-strict requirements that capture size exactly match the native resolution. We want libvidcap to honor any reasonable framesize request, regardless of the native resolution. That and the colorspace changes are in their infancy, so pushing them to trunk are slightly premature.

We welcome any suggestions, input, and help.

Cheers,
Bill