-
Audio Track Extraction?
Is there a program out there that will allow me to feed it a file (an mp3 or even a video clip) and have it analyze the data and split off each audio source into its own track?
I tried to do this a few years ago for musical training. I was trying to strip out just the bass from a song so I could practice as main guitarist. I found nothing that could accomplish it back then. I posted on some forums and was met with wailing and gnashing of teeth. “That’s way too difficult!” and “Nothing like that exists!” and “It’s practically impossible to accomplish!” and so on.
However, I happened to see some random crime documentary over the weekend and on it they were viewing crappy black-and-white security camera footage and the IT forensics guy clicked a magical Easy Button and BAM! the audio split from one track into about five tracks. He went through each one. Paraphrased: “This is noise from outside the store, we’ll mute that. This is the clerk gibbering madly, we’ll mute that. Aha! Here’s the nefarious villain speaking. Let’s cross-reference this voice pattern with the voice pattern from another robbery and see if it’s the same ne’er-do-well.”
Now, this wasn’t CSI or anything apparently fake. What program are forensics people using?
Thanks
-
4 Replies
-
Here’s some interesting insight… I didn’t read much into this, but maybe getting close. I do some audio restoration and remember finding a software years ago that was about $10K US to isolate sounds. But tech has come a long way. I searched “forensic audio”.
One of the links…
https://www.cs.helsinki.fi/u/ahyvarin/whatisica.shtmlMaybe this…
https://speechpro-usa.com/product/forensic_analysis/fawThis…
https://sourceforge.net/projects/isse/Steve
-
Hello Roy and welcome to the Cow Audio Forum,
What exactly are you trying to do now?
Regards,
Ty Ford
Cow Audio Forum LeaderWant better production audio?: Ty Ford’s Audio Bootcamp Field Guide
Ty Ford Blog: Ty Ford’s Blog -
So think of sound like paint. Know how when you mix paint, it’s next to impossible to unmix it back to its component colors? Start there. Now some specific sounds are “different” enough to be able to isolate and manipulate. So it’s not hard to filter out, say, a high frequency screech from a recording of a kettle drum. They exist in distinct frequency bands. It’s WAY harder to separate one voice from many voices, because all voices are relatively alike in frequency and volume – at least compared to a Kettle drum and a screech.
Now the big hassle. Almost no “real world” sound is simple and pure and therefore easily isolatable. When you say “hello” you don’t do it at a single frequency, but rather create a host of complex vibrations with fundamental tones and overtones and harmonic tones galore. This means that to isolate YOUR voice like on a TV crime show, some technology has to figure out how to separate just your mess of sounds from all the other competing messes of sounds – even though there isn’t much difference mess to mess.
For the person Making the Show, however, there’s a vastly easier and cheaper way. Just start out with the required sounds as isolated elements, mix them together temporarily to get a realistic sonic stew – then remove the elements one at a time to get back to one clean track. And that’s precisely what you gave seen and heard on a zillion TV shows and movies.
Forensic audio can isolate what is “isolate-able” -but if the sound is completelt mixed up like a can of green paint that started out as equal partsf yellow and blue – separating back to the component colors is just not going to happen.Know someone who teaches video editing in elementary school, high school or college? Tell them to check out http://www.StartEditingNow.com – video editing curriculum complete with licensed practice content.
-
Nice one, Bill!
There is Celemony’s Melodyne. It’s not about extraction, it’s about pitch modification, but you can “erase” individual notes and overtones.
https://www.celemony.com/en/start
I’ve found it very interesting, but either I need to take a sabbatical to learn its depths (as one might learn Photoshop) or I have reached its limitations.
iZotope’s RX3 Advanced is about noise reduction. Again, not file extraction. Once you get rid of it, it’s gone.
What I have been able to do occasionally, is shocking! 🙂 If you have the time, please stick your face into these and report back to us with your findings!
Regards,
Ty Ford
Cow Audio Forum LeaderWant better production audio?: Ty Ford’s Audio Bootcamp Field Guide
Ty Ford Blog: Ty Ford’s Blog
Reply to this Discussion! Login or Sign Up