Synching related but not identical audio tracks

[Peter Grad]

I agree, I acquired that app, and SoundHound, when both of them first came out and was stunned at how cool they were. Even more amazing, as great as they were just a few years ago, what used to take about 30 or 40 seconds now takes as little as four or five seconds to identify a song. You can even hum a tune, imperfectly, and SoundHound will nail it. I even tried it with an old Asian girlfriend of mine and her mom sang a song in Chinese, and SoundHound nailed it…

[Bret Williams]

Perhaps the old show “Name that tune” was just a front to recruit shazam employees that actually listen to a small recording! Must be it.

[Noah Kadner]

Shazam is actually incredibly complex but it relies on the fact that mathematically speaking every song has a unique DNA that can be quantified and matched from a database- just like your DNA and your fingerprint. How it works:

https://gizmodo.com/5647458/how-shazam-works-to-identify-nearly-every-song-you-throw-at-it

Which is not to say that audio sync shouldn’t be made smarter and able to account for substantially similar, though not identical waveforms. But at this point it doesn’t.

Noah

FCPWORKS – FCPX Workflow
FCP eXchange – FCPX Workshops

[Mark Suszko]

I’m trying to follow this and have to ask: has everybody gotten so lazy from auto audio syncing that they’ve forgotten how to jam-sync free-run time of day time code into multiple cameras? Why can’t you sync the wild clips by their time code? Or am I missing something?

[Peter Grad]

Here is what you may have missed:
1) I am a beginner wirh Final Cut and am notmfamiliar with time synching.
2) I may have been unclear, but I am using one audio-only soundtrack, composed with Logic Pro, and then attempting to match it up with four video clips, each recorded separately, each representing different melody lines – the lede and three backup harmonies. I am using a Canon Vixia HF R60, which I dont believe offers a time stamp, but perhaps i am wrong. Each of those video tracks was recorded separately, not simultaneously, so I am not certain your good suggestion of time synching would work in this case anyway since the time stamps would all be different.

[Noah Kadner]

Sorry for the detour- bottom line is no auto-sync for you.

Noah

FCPWORKS – FCPX Workflow
FCP eXchange – FCPX Workshops

[Mark Suszko]

Thanks for clarifying, Peter. In this situation, what I would suggest for the next time is to always record the full song for each take of each part, no matter if it’s mostly silence with just a simple punch-in of the fourth tenor singing one little “ooh, yeah” or whatever in one chorus section. This gives you equal length “stems”. If every take is a consistent length from a consistent starting point, everything will line up for you naturally, you can align automatically OR manually, to the starting click… then your only task is to manually go into each track and divide it up into the parts that are audible, and their relative mixing levels. Use a click track or visual sync point at the start, before the song, an you can edit that out of the final mix.

[Peter Grad]

Thanks Mark, I’ll have to look into the use of click tracks…

[Peter Grad]

Well, we’ve got the precision of Shazaam and SoundHound to match songs perfectly, someone needs to apply fuzzy logic to an app that will match up approximate sounds one of these days…

[Noah Kadner]

Also SMPTE timecode may be old school but it works 100% of the time…

Noah

FCPWORKS – FCPX Workflow
FCP eXchange – FCPX Workshops