[LAU] OT - Audio and Video Synch

Len Ovens len at ovenwerks.net
Fri Oct 18 22:38:26 UTC 2013


On Fri, 18 Oct 2013, James Stone wrote:

> I am wanting to get some audio and video recorded - sample accurately
> - so there is minimal delay between audio and video, and no drift. Can
> any format do this? Do you have recommendations about how to record a
> file which will play back with perfect audio and video sync on all
> players (file size is not an issue)?

The same player might have non-sync video/audio depending on the audio 
output path. One would expect the player to compensate for the delays 
cause by decoding, but then the audio may go straight to the alsa device 
or through pulse which adds latency or through jack (where you can set the 
latency so one time might have a different delay than another) or through 
gs first then through who knows what. Some of the sound servers may have 
variable latency depending on cpu load too.

That is not even touching the video chain (outside of my knowledge), but 
you get the idea. On any reasonably fast computer one can set up an 
audio/video chain that will work in sync, but change any one part of that 
chain and sync may be no more. So if you are looking from a distribution 
POV, you have no control of what the audience will play your video on. If 
you are using it in an institution, then all the machines may be the same 
as well as the OS/SW and so there is a better chance. Keeping the decoding 
work as low as possible and using sample rates and frame rates that match 
the sound serve/audio IF and work as close to the GPU setup as possible 
reduces things like resampling/resizing and reduces cpu load and brings 
different speed HW closer to matching. For example, if your audio is using 
opus as an codec, then 48k sample rate in the sound server/audio IF will 
result in no resample to match the codec ouput. But if the audio card is 
set to 44.1k then the audio must be resampled. However, many of the videos 
available on the Inet have audio at 44.1k and so running the audio IF at 
48k would mean a resample step for them.

So it would seem it would be impossible to make sure that audio and video 
are always in sync on every machine. The recording and encoding sw can 
shift the audio with respect to the video. Then check it out on the 
machine/OS that most of your target audience uses.

I don't know how the windows audio server works, however, the intel HDA 
audio interface assumes 48k (or 96k or 192k... though even at these speeds 
it seems the bus runs at 48k with just more parallel data). So I would 
guess the default windows setup would be 48k. I don't really know though.

--
Len Ovens
www.ovenwerks.net



More information about the Linux-audio-user mailing list