How do I extract same part from two different V of song?

So, I have two different version of a song. Two songs are exactly same, but was sang with different language.
Maybe you can think of some Disney songs in different for example, they have exactly same instruments, but vocals.

What I am trying to do is to extract this instrument part from the song, (and not using vocal remover)
I tried ‘subtracting’ one from another, but soon realized I was removing instruments instead of removing vocals.
Since I got only vocal part, I thought subtracting vocal part from mixture of two version would remove vocals.
In fact, I could remove only one language from the song.
I suppose it is because acapella version I got from first method has one language reversed, thus cancelling only one version.

I thought comparing both version and extracting same part is easy, but it doesn’t seem so.
Is there easier way to compare two version and extracting only duplicated part?

Probably not. You have two different songs and I don’t mean they’re in different languages. I mean the language change causes the music compression algorithm to change everything, not just the voices. So the music may be the same to your ears, but it’s usually different enough to make it almost impossible to “isolate” or process individual parts.

Usually, the only way to remix a song is to have access to the original multi-track recordings.

And some of them are available on-line.

The latest 2.1.2 Audacity has vocal removal and isolation, different from earlier versions.

Maybe there’s something in there.