These are recordings of game audio, so the two identical sounds are actually sound files somewhere, and pretty much guaranteed to be identical (the audio that is played is positionless, and unaffected by conditions within the game).
There is a clue buried in the game audio that I need to extract, but I want the absolutely cleanest copy of the clue I can possibly obtain. Others have already pulled out the clue in a fuzzy, distorted form, and have managed to find nothing with it.
I would like to use these two recordings to keep the identical sound between the two and toss everything that’s not identical.
I know about phase inversion to do the exact opposite, but I don’t need to cancel out the identical sounds, I need to keep only those.
Phase Inversion is the easy one. Straight, plain arithmetic and it can give perfect results. Everyone always assumes you should be able to simply reverse the process and get isolation. Not so far. Even if that’s done, it’s done by acoustic tricks and digital slight-of-hand, not perfect arithmetic. It may not give very good results.
It is sometimes possible (but perhaps not legal) to deconstruct the game files and digitally extract (rip) the audio data. Of course you shouldn’t attempt this if the copyright/license terms make it illegal to do so in your country, but if it is legal then you could google for “how to rip game audio” or similar search terms.
Yeah, I’ve seen that option, but of the eight or so choices, I’m not sure which one to pick. I tried three or four, but got bad results each time. I don’t know what each option actually does, so I don’t know if I’m just doing it wrong or I don’t have things lined up correctly.
Recorded them via nVidia ShadowPlay. The audio is in AAC format, which I understand IS lossy.
I’ve seen that, but it keeps referring to things like ‘vocals’ and ‘panning’ and ‘center’. I don’t have vocals, and I don’t know what the other two things are. (Well, I have educated guesses as to what they are, but not really sure guesses.)
I don’t think you need to worry about that as it is very unlikely to give useful results for what you are trying to do, but just for academic interest:
The “Vocal Reduction and Isolation” effect works (when it works) by looking at where sounds appear to be coming from in the stereo field. That is, it looks for sounds that appear to be coming from the left, the right or the middle (this is called the “pan position”). If the sound that you want to keep (or remove) is coming from dead centre, and all other sounds are coming from the left or the right, then the effect allows you to keep (or remove) that “centre panned” sound.
From your description I’m guessing that it’s unlikely that the sound that you want to isolate is in the middle and all other sounds to the left and right, in which case this effect will not be useful for this task.
Is there a way for me to use phase inversion to keep all the sound I want to get rid of, then use that sound to filter out what I don’t want from the original file? (When I phase invert, the remaining sound is quite muted, but it does sound as if I eliminate the sound I want to keep almost perfectly.)
then use that sound to filter out what I don’t want from the original file?
You would think, right? Did you catch when you got your “clean” result at the end of flipping and cancellation, it was mono and not stereo? That’s what kills you. There’s no way to get back to the stereo show, and you would need to do that for any of those other tricks to work.
Some very talented and energetic people have tried to make this work. The closest we ever got was that Reduction and Isolation software.
Is there a way for me to use phase inversion to keep all the sound I want to get rid of, then use that sound to filter out what I don’t want from the original file?
Moleculor, how is your algebra?
Lets say you have 2 sound files:
File A = X + Y
File B = X + Z
X is the sound in both files that you want to keep. Y & Z are the unwanted sounds mixed-in…
If you play around with the math, you’ll discover there’s no way to eliminate both Y & Z, leaving X. However, you can subtract the two expressions to eliminate X. (In the audio editor, subtract by inverting one file and mixing.)
Your intuition might tell you that you can take the result (Y-Z or Z-Y) which does not contain X, and do some further math to get X, but your intuition is wrong…