Replacing existing audio

Hi, I'm an audio guy, and am getting into video editing. A friend of mind recently shot some video of me and another friend while we were recording our podcast, and he is going to give me the raw footage on DVD very soon.

I want to replace the audio that the onboard mic captured with the audio I recorded with my good mics, etc.... I'll probably have to resample to 48K from 44.1, but that's nothing.

The question is, how difficult is it to replace the audio that accompanies the video footage, with my audio, and how hard is it to line it up. Can this be done?

Thanks,
Rich...

You can easily separate out the audio from a video track in Kdenlive. Just right-click on the track and hit 'Split Audio'. The two clips will still be grouped, so you may want to 'ungroup' them. Then you can mute that or turn it's volume down or delete it. Keeping it around rather than deleting it can help you place the replacement audio, especially if the two are the same material, just differently mic'd. Importing the new audio clip should be no problem, then just drag it to an audio track. Syncing the two up is more a matter of how complex and piecemeal the video and audio are. If you just shot one take and also left the audio recorder going, then you probably just have to do it once - you just zoom way in on the timeline and move the audio track back and forth manually and check the sound sync. You can move a track pretty much down to the nth degree.

The more cuts there are in the material, well, you'll have to do this manually for every clip. I've done this with offboard mics and a Zoom audio recorder, and it's not a big deal, again, depending on the number of clips and cuts.

With AVCHD playback being kind of jerky, it can be hard to tell within Kdenlive itself if audio is actually synced. In that case you need to render a short bit and play it in a good media player to check.

Once the two are aligned, you can link the video back to the audio by ctrl selecting both, right click, and hit 'group clips'. You can't re-integrate the audio back onto the video track, but if the tracks are grouped, moving one moves the other.

The biggest problem is editing after the sync - edit the video clip, and the audio track is not effected, even if the two are grouped. You can still edit, it's just a little trickier. In order to edit this easily, it would be nice to be able to truly unite the video and audio back together, and I just don't see that that's possible directly. You could do an intermediate render of the whole thing to lossless codec and then edit that I guess.

> You can easily separate out the audio from a video track in Kdenlive.
> Just right-click on the track and hit 'Split Audio'. The two clips will
> still be grouped, so you may want to 'ungroup' them. Then you can mute
> that or turn it's volume down or delete it.

or you can add the "mute" effect to the video clip

OK, I like what I'm hearing so far. Cool. Shouldn't be too much work. Any other suggestions??

What non-lossy codecs does kdenlive support?

HuffYUV, FFV1, DNxHD, FLAC

I am adding this issue to the to-do for tutorials:
http://www.kdenlive.org/forum/possible-tutorial-subjects

It can be a simple and interesting tutorial.

See that? I think of the good questions. :-)

Still didn't get the raw video footage in the mail yet.

Please correct me if I'm wrong, but I think that there's a limitation to syncing audio to video (imposed by mlt?):

As far as I know sample accurate aligning of audio to video is not possible:
audio can only be shifted in quanta equal to the video framerate.

Well that would be a problem.

True, but how accurate do you think humans are at perceiving a/v sync error? Do you believe most humans can perceive an a/v sync error less than 40ms PAL or 33ms NTSC? Maybe you can add a SoX or LADSPA delay filter through a kdenlive effects XML file to address sub-frame accuracy. It would be great if you can experiment with that and contribute an effect XML that works for you.

In fact, on average the error is 1/2 the framelength, that is, 20 ms pal, 17 ms. NTSC.

I do believe though that *some* people would be able to tell the lag/difference (I used to play video games quite a lot, thank-you :-), but delaying would be interessting to experiment with.

Syncing is not the only issue:
The main problem would be if you try to cut out small parts of audio in the middle and paste both remaining parts back together: because it cannot be done in a sample-accurate way (e.g. at a zero transition), you are left with audible artifacts at the places where you cut.

Very good point, Stefan. Audio editing to the precision of a sample is not something we can offer in kdenlive anytime soon. It might easier and better to offer Jack integration including transport sync.

Pages