> I'm having a problem with the audio sync on a rendered video. This is a > video of a rock band playing. I have used the original audio from the > cam (720P HDV, MPEGTS, 48.1khz audio) to sync up a multi-track recording > of the same session (WAV, 48.1khz audio). When reviewing the audio > waveform in Cinelerra, the new and old audio tracks align perfectly: > http://content.serveftp.net/video/sync_audioalignment.png
I'll have to take your word for it; the waveforms as presented are rather difficult to use to verify sync. The other thing to note that at least at the scale in the above screen grab, two tracks can *look* in sync but be far enough out to make the sync mismatch audible. Hmm, I assume you mean 48 kHz. One question before we start. Was your multitrack and camera synced to the same clock during the recording session (be it word clock, smtpe etc)? If not, the reality is that the sampling of your multitrack and that of the camera audio will differ slightly. The difference might not be large, but even over 4 minutes I would expect a detectable sync problem towards the end. The only way to resolve this with unlocked sources is to resample the audio from the multitrack in such a way that you correct for the subtle clock differences of the gear used. I routinely do this kind of thing and wrote a program to do this resampling. You can check it out at http://www.physics.adelaide.edu.au/~jwoithe/audiosync-0.4.tgz It's quick and dirty but does the job for me. In the context of your job you would pick to points in the recordings which are easily identifiable in both waveforms - preferably at the start and end of the audio. You then work out the exact sample number that this event corresponds to. While this can be easy for direct audio (from the multitrack in your case) room reverb can make picking events difficult in the camera audio (unless you piped the direct sound into the camera as well). Usually I pick things like the snare hits, for obvious reasons. These sample numbers are then used by audiosync to resample the multitrack recordings so those events line up exactly. So long as you've chosen accurate samples you will end up with very good sync over the long term (for example, using this method I often achieve synchronisation to within 10 samples over a period of 90 minutes). As far as the software you need to work out the sample number of these events I use audacity. I set it to work directly with the audio files since it loads the files much quicker (although with only 4 minutes of audio you are unlikely to have problems either way). You can set the location display to be in samples using an item about half-way down the "View" menu. The zoom buttons are then your friend. > Also, when playing back the video in the compositor, I hear no evidence > of a sync problem. I can turn on both cam and externally recorded > stereo tracks and listen the whole way through the four minute recording > without any reverb or noticeable sync issues. However, after rendering > the files to a final format, I get inconsistent results in xine and > mplayer. This seems rather strange. If there's a sync problem you should hear it in cinelerra unless for some odd reason the tracks aren't always being mixed. I must admit that I do *all* my audio mixing in ardour (I export the camera audio to WAV, import that to ardour and then sync off that), so have no experience doing multitrack audio mixes in cinelerra. > sync.mov: Xine shows sync problems, mplayer no problems > sync.mpg: Xine shows no sync problems, mplayer shows problems That sounds bizarre. While you can get sync problems in certain situations with different AV container formats this would cause the entire audio track to drift out of sync. As I understand it from your opening statements, "sync problem" here refers to an audible echo indicating one source track being out of sync with the other. Regards jonathan _______________________________________________ Cinelerra mailing list [email protected] https://init.linpro.no/mailman/skolelinux.no/listinfo/cinelerra
