[FFmpeg-user] Comments about audio sync invited

[James Linder]

On 11/07/2013, at 2:49 PM, Oliver Fromme wrote:

> James Linder wrote:
>> 
>> On 10/07/2013, at 10:19 PM, Nicolas George wrote:
>> 
>>> Le duodi 22 messidor, an CCXXI, James Linder a écrit :
>>>> [dvr] /home/dvr [132]% ./ffmpeg -f pulse -i default -f v4l2 -i /dev/video0 -vtag divx -s 384x288 -r pal -q:v 5 -f tee  -map 0:a -map 1:v Wed.avi | cat >Wed1.avi
>>> 
>>> The tee muxer in this command line does not make any sense. The cat is
>>> completely useless too. 
>> [...]
>> Why is my cat completely useless?
> 
> Because the following two commands produce the same result:
> 
> $ ffmpeg ... | cat > foo.avi
> $ ffmpeg ... > foo.avi
> 
> The difference is that the first has an additional overhead
> of one execve() (probably insignificant in this case) and
> of one pipe and one process that have to handle the output
> data.  The latter might be significant because it usually
> involves at least two in-memory copy operations, depending
> on the implementations of pipes and cat in your operating
> system.
> 
> The above is a typical "useless use of cat".  It should be
> avoided.

Methinks that you do not understand what cat was being used for (to determine if system load was an issue using a |pipe|)

What is clear after building git head

Stuttering is no longer an issue
Audio sync is manageable (I will look at hacking the code to force same-sort timestamps) but this is really a case of using the latest.
Many thanks to Carl: the questions and comments were very helpful.
James