Best Cues for Audio Synchronization (Manual)

[uart]

Sometimes we have to resort to manual adjustment of audio advance/delay (and watching the result) to try to tweak the synchronization of a badly sync'ed video. I'm not asking about the software or procedures as such, but more about what cues you use when watching the video to detect when the audio sync is most optimal.

Also sometimes just the one audio offset is not really enough to fully correct the entire piece, but maybe it's close enough if it's chosen carefully. I've read that in general audio advance is perceived as more annoying than audio delay so I suppose that should be taken into consideration as well. Anyway I'm just interested in other users experiences. How they handle the problem of manual audio synchronization, what cues they find most useful when checking for good sync and how they handle compromises when it's required.

BTW. My motivation here is that I'm in the process of archiving some recorded TV (mpeg2 transport stream) to x264 codec mpeg4's. The results are pretty good but in some cases I just have to resort to manually tweaking the audio delay to get a good result, and it's often a pretty tedious and annoying process.

Thanks in advance for any input.

[poisondeathray]

You usually have to process in segments. often there is a transmission error or drop out, and you have to cut out that bad piece in order to get it sync properly. Often there are visual cues as to where the problem areas are (e.g. pixellation, distorted picture)

avidemux can shift on the fly - so you can enter a +/- audio shift number, play, and adjust, play etc.. , still tedious but it might be a bit easier than what you're currently doing. So you would process segments until each was in sync, then join them later - instead of picking some intermediate value for a whole recording which is a bit off for everything

Obvious things like doors slamming, close ups of actors talking are good cues to follow

Sometimes it's easier to do in a dedicated editor like vegas or premiere (a non linear editor with a timeline) , so you can see the waveform of the audio, & nudge/adjust

[AlanHK]

With MPC you can adjust playback audio delay simply by tapping the + and - on the number pad.
When you get an acceptable figure, you can demux and apply the delay, remux.

Another way is to use Avisynth, with script at this page:
http://avisynth.org/mediawiki/FindAudioSyncScript

In the script you can set this at half, quarter whatever speed, so you can judge the delay as fine as you want -- usually I select a short segment with a sharp noise -- a cup being put on a table, eg. I find it too confusing to try to sync speech to a mouth -- you know when it's wrong but it's very hard to tell which way to adjust.

You should check at least two sync points, near the beginning and end, to see if the delay is constant. If not, you can stretch it to fit using Audacity.

[uart]

Good info there guys. Thanks.

avidemux can shift on the fly - so you can enter a +/- audio shift number, play, and adjust, play etc.. , still tedious but it might be a bit easier than what you're currently doing.

Ok that's good. Actually avidemux is the app I'm currently using but I hadn't previously noticed the "on the fly" adjustment. I'll give that a try.

I find it too confusing to try to sync speech to a mouth -- you know when it's wrong but it's very hard to tell which way to adjust.

Yes that's exactly the problem I've been having Alan. I can probably get it to within +/- 100ms just looking at the lips but I find it hard to be sure of just where the best value lays.

BTW that brings me to another question. How close is close enough? Does anyone have a good feel for what amount of sync error is acceptable. Lets say for example we are considering an audience who wont be critically viewing on the lookout for sync errors or anything, but for "casual" viewers who we would hope will not notice or be annoyed by any sync problems.

What are we looking for? +/ 50ms? +/-100ms? ...?