On Tue, 24 Jan 2012 14:48:03 -0500, Justin Ruggles <[email protected]> wrote: > On 01/19/2012 03:01 PM, Justin Ruggles wrote: > > > According to unofficial documentation, the video rate is locked to the audio > > sample rate. This results in proper synchronization of audio and video > > timestamps. > > > One problem with this patch is that it changes the time_base after > read_header(), where the stream is created. > > In all the samples we have, there is an audio packet before the first > video packet, so the simple solution would be to store parameters in the > private context and create the video stream when the first video packet > is encountered. If at that point, we have read an audio packet, we can > set the correct time_base, otherwise we have to use the default. > > A more difficult solution would be to buffer video packets until we get > an audio packet (or a certain number is reached, indicating we probably > don't have an audio stream). I actually implemented this at one point, > but I couldn't test it since none of the samples in our archive need it... >
How many samples does this format actually have? It's an obscure game format, so I'd say implement whatever is easier and don't waste too much time on this. -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
