si1ver
29th August 2003, 05:05 AM
hi,
what are good codecs for vj loops? I'm using a mixer i've built in pd/GEM and i am struggling with codec choice. I've resigned myself to using 320x240 res clips in order to mix 3/4 layers of video at around or above 20fps (how many layers of video at 20fps should i be able to mix with 12' G4??)- but my lap top struggles when playing the clips in reverse. I presume this is something todo with temporal compression/difference frames that most video codecs utilise to reduce file size and while the codecs decompression algorithms work with minimal computational overhead when playing forwards the same is not true of clips playing in reverse- all those nasty difference frames. Do I need to make every frame a keyframe? Or should I use photo-jpeg compression, storing each frame as JPEG as suggested in the Jitter manual.
Also I was wondering if any codecs surpport alpha transparency- so a clip could be saved with areas of transparency already defined (using keying on final cut for instance) and remove the need to key live, as this is quite a big convolution to be done in realtime. I know JPEG does'nt do transparency of this kind but PNG will and quicktime provides an option for formatting video clips as a stream of PNGS but PNG is not really suited to photo realisitic content (alot of my clips are of more natural scenes). I imagine the table bassed look up of PNG files is not the easiest thing for a computer to do. Obviously decoding one 320x240 PNG file is easy, but what about doing almost 100 each second + all the other stuff.
this codec dilema has been a concern of mine for several months now and i am a firm believer that good codec choice is crucial to fluid high frame rate playback. to this end i have been expirmenting but with no definte conclusive evidence (appart from avoiding MPEG4 with it's oodles of compression). i would be grateful for any insights.
cheers tim.
what are good codecs for vj loops? I'm using a mixer i've built in pd/GEM and i am struggling with codec choice. I've resigned myself to using 320x240 res clips in order to mix 3/4 layers of video at around or above 20fps (how many layers of video at 20fps should i be able to mix with 12' G4??)- but my lap top struggles when playing the clips in reverse. I presume this is something todo with temporal compression/difference frames that most video codecs utilise to reduce file size and while the codecs decompression algorithms work with minimal computational overhead when playing forwards the same is not true of clips playing in reverse- all those nasty difference frames. Do I need to make every frame a keyframe? Or should I use photo-jpeg compression, storing each frame as JPEG as suggested in the Jitter manual.
Also I was wondering if any codecs surpport alpha transparency- so a clip could be saved with areas of transparency already defined (using keying on final cut for instance) and remove the need to key live, as this is quite a big convolution to be done in realtime. I know JPEG does'nt do transparency of this kind but PNG will and quicktime provides an option for formatting video clips as a stream of PNGS but PNG is not really suited to photo realisitic content (alot of my clips are of more natural scenes). I imagine the table bassed look up of PNG files is not the easiest thing for a computer to do. Obviously decoding one 320x240 PNG file is easy, but what about doing almost 100 each second + all the other stuff.
this codec dilema has been a concern of mine for several months now and i am a firm believer that good codec choice is crucial to fluid high frame rate playback. to this end i have been expirmenting but with no definte conclusive evidence (appart from avoiding MPEG4 with it's oodles of compression). i would be grateful for any insights.
cheers tim.