obtaining background audio in a video clip

[muffinman123]

the video clip has several sound effects and some loud talking. what is the software to do this and how hard is it to do it?

I know the track cannot be completely recovered and the best I can do is to diminish the volume of the foreground sounds, but I really want to listen to it.

it's the soul calibur 5 ezio's venice rooftop remix, at best I can only get 2 half pieces because it's cut off in the middle, but I mind, I just want to get these 2 parts.
http://www.youtube.com/watch?v=dBVdXToUvII&t=0m51s

[hech54]

It's a video game.




Easier to wait for the soundtrack:
http://en.wikipedia.org/wiki/Discography_of_the_Soul_series#Soulcalibur_II_Original_Soundtrack

[muffinman123]

that's true, but I want to learn how to do stuff like this. it's for the fun of it.

[Cornucopia]

If you want to learn "how to do stuff like ths", you're going about it the wrong way.

Extracting parts from a homogenous whole is some of the most difficult things in audio or video. You don't start with the harder stuff and learn and graduate to the easier stuff, it's the other way around.

What you could do in this particular case is play (and capture) the game with one of the elements missing. Often in games you can choose to mute the music track or the effects track.
If you can play the game NEARLY IDENTICALLY multiple times, you can cap all variations of Vox+Music+Effects tracks and, after some SERIOUS timeline synchronization, could SUBTRACT the tracks you don't want.

For example: if you have a choice of V+M, V+FX or V+M+FX, you begin with V+M+FX and subtract V+M to get FX. Then you take the V+M+FX and subtract the V+FX to get the M. Then you take the M and the FX, combining them to get a M+FX track. Then you take the V+M+FX and subtract the M+FX, getting a V track!

But this ONLY works at all if you have identical LEVELS, TIMES, and PHASES. With a repeatable game with soundfile playback, it's POSSIBLE to get to that point, but extremely difficult and time-consuming (unless you game has a feature that "plays back your moves" and you play it back multiple times with the different tracks enabled - then it's easy).

If you aren't talking about gameplay, then it gets even more difficult and you have to resort to Filtering, DynamicRange adjustments, and NoiseReduction tools, all of which work generally and incompletely at best...

Scott

[hech54]

that's true, but I want to learn how to do stuff like this. it's for the fun of it.

Scott is right. What you are trying to do only happens in movies and on television like Navy CIS. It doesn't work like that in the real world.

[muffinman123]

If you aren't talking about gameplay, then it gets even more difficult and you have to resort to Filtering, DynamicRange adjustments, and NoiseReduction tools, all of which work generally and incompletely at best

I know the recovered audio will be incomplete and will carry left parts from original sound effects as garbles or some sort of noise, but I just want to try, and I want to know what kind of software can do it.

just list some software with the function you listed, and I will probably try a bit. maybe I will give up, but at least I know I tried.

[Cornucopia]

Any decent multi-track DAW will work, as they are all sample-based (not frame-based like video, and that's important), have room to maneuver tracks and adjust relative levels, and can apply varying levels of processing (filters, etc as prev. mentioned). This includes: Ableton, Sonar, Logic Pro, Sequoia/Samplitude, ProTools, Reaper, Nuendo, Audacity, probably others...

It's not so much which app, as the process (and knowledge of how to manipulate). This requires an understanding of signal theory. I won't go into "steps", because it would fill a book and there are already good books out there. Much of those steps greatly depends on what is happening in the composition of the signal - that is different every time, so you have to know what and how to change things to get your desired output (if that's even possible - many times it isn't).

And NO, there really aren't any tools/plugins that automate this kind of adjustment. There are a few tools that help facilitate and speed up a section of a process, but even those are $$$$$. I don't even have access to them (probably wouldn't unless I was in Hollywood, Gov't/Military, or won the lottery).

Scott

[muffinman123]

audacity only has fourier transform for analysis.

damn it, even with basic electrical and image signal analysis, audio is different. it would be nice to be able to listen to the individual frequency separately and then filter out the ones that I don't want. is there a tool that decompose the frequency into separate tracks for comparison?