0
$\begingroup$

I use the Blender VSE for overlaying live video footage from a camcorder (with noisy audio from the camcorder) with good quality audio straight from the mixer, captured at a live event.

I need to manually align the two audio tracks, after which I remove the camcorders' original audio track.

However, after 10 minutes, there's a lag between the two. It slight, but annoying. After 20 minutes, it's symply out-of-sync.

Q: how can I avoid this difference?

Tried so far: -The problem appears in both the preview and in the rendering. (so it's not just down to the av-sync button) - The frame rate of Blender's rendering and the video match: 25 fps - The recording frequencies of both the audio-file and the video file are 48000 Hz - The video has a constant bit-rate

Camcorder track (video):

Stream #0:0: Video: h264 (Constrained Baseline), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 25 fps, 25 tbr, 1k tbn, 50 tbc (default)

Audio track:

Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, 2 channels, s16, 1536 kb/s
$\endgroup$
  • $\begingroup$ Small sampling differences add up quickly sorry. $\endgroup$ – 3pointedit Nov 18 '17 at 23:32
  • $\begingroup$ @3pointedit Thanks for the comment! With 'sorry' you mean: there's nothing we can do about this? $\endgroup$ – Ideogram Nov 19 '17 at 9:05
  • $\begingroup$ Yup... sorry. Best option would be to resample the audio but blender can't do that, also you don't know if the fluctuation is variable. You could try changing the Audio Strips Pitch value, but I don't think that it will give you a fine enough change. And the pitch will alter of course. $\endgroup$ – 3pointedit Nov 19 '17 at 11:30
  • $\begingroup$ @poinedit If you say, "Best option would be resampling". Sure, can do that in audacity. But what should I resample it into? Current: wav @ 48kHz. Or is it maybe the video that should be resampled. Your comments are already very much appreciated! $\endgroup$ – Ideogram Nov 20 '17 at 7:30
  • 1
    $\begingroup$ Looking at the codec parameters it looks correct. Hmmm, when we shoot stuff for TV we sync long recordings so that they do not drift from each other. This involves using a video sync lead between recording devices. You may have to manually break the audio and resync to video at regular periods, sorry $\endgroup$ – 3pointedit Nov 20 '17 at 8:35
1
$\begingroup$

Bad sync between recording devices happens for many reasons, check out the wiki article

Digital or analog audio video streams or video files usually contain some sort of explicit AV-sync timing, either in the form of interleaved video and audio data or by explicit relative time-stamping of data. The processing of data must respect the relative data timing by e.g. stretching between or interpolation of received data. If the processing does not respect the AV-sync error, it will increase whenever data gets lost because of transmission errors or because of missing or mis-timed processing.

As the audio has drifted away from your pictures without a common reference I would suggest looking for easily identifiable sync points in both recordings. Split the audio at that point and drag it back to match the video. Depending on how fast the drift occurs will indicate how many splits you require. It could be one every few minutes if you're unlucky.

You can try to apply a minimal amount of pitch control on the audio strip. This will speed up the playback of the media by a certain amount, but its hard to be accurate as the wave form won't reflect the change, unless you render it out and re-import it. Use Blender's Audio Mixdown function to achieve this.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.