Trevor Boicey (tboicey@xxxxxxx): > I'm fine with this, except I run into loading issues, where the > encoder running on /dev/video will drop frames, and then the > characters in the captions for those frames are also lost. This leads > to garbled captions and unreadable stretches of text. There are other reasons this might happen with the ccdecoder code. For one, the filter used is not very good, and we need to improve it. We had alot of problems originally with that code because it would screw up the multiple CC channels, and get confused by XDS, and not look in both fields, etc. The code was based on some stuff hacked together by Nathan Laredo. > For background, I am attempting to capture low-quality video from > /dev/videoX as well as captions from /dev/vbiX. The video format can > be anything that is commonly supported (ie: Windows Media Player). So > far I am using ffmpegrec, although I have had mixed success with > ffmpeg and mp1e. All suggestions appreciated. Tx. Captions are difficult to capture just as a stream of text because there is alot of rendering information: position on the screen, how long to show it for, etc, that should be taken into account for it to make sense. Take a look at the tvtime sources: http://tvtime.sourceforge.net/ In CVS we have an improved version of Nathan's code, with still some bugs, and an attempt at a compliant cc renderer. There is still alot of work to do though, but we do intend to implement rendering per the spec. -- Billy Biggs vektor@xxxxxxxxxxxx