So guys I’m gonna be as specific as possible:
This band I like just released a new album and a special version of it containing a CD with the songs and another CD with just the instrumentals.
The first thing that crossed my mind was “they just provided us an easy way for us to isolate the singer’s voice acapella!”. Oh, how totally wrong I was.
After ripping it to my computer, I loaded audacity to get it going. My first surprise was: They mastered the songs in a way that the instumental version of all songs are slightly longer than the original songs, meaning the total time of the songs with the voice is 40:50. The instrumental version is 40:55. My first thought was “it’s probably some extra silence in the beginning/ending of each song.” But it’s not.
Take for example the first track. I perfectly aligned the start of both versions (with and without voice) but as you can see the ending doesn’t match:
Beginning:http://i.imgur.com/rpY2Om4.png
Ending:http://i.imgur.com/Twni3c7.png
Entire track:http://i.imgur.com/ASFmzwU.png
The instrumental track is about 00.700 seconds longer. And no sound was added to it, it seems as if it was stretched, but only enough so there is no pitch difference. If you play it together, you can see it slowly gets unaligned, and by the ending it’s all out of sync. Every track is like this, all the instrumetal ones are “stretched”.
You guys know where I’m going? If the tracks don’t align, I can’t use “invert” in one of them to get just the voice. This is so frustrating. The first solution I tried was using “effect → change tempo” to “shrink” the instrumental track, but still no success. Somehow I still can’t set the correct duration.
I know this is a long shot, but is there a way to make audacity automatically set the correct tempo so they will align?
Thank you all in advance.
I would try Effect > Change Speed anyway. That tool has a place to post the old and new times and it will try to resolve them.
Koz
I tried it too, but no success. I think I need a “fine tune” tool. you see, for example, “change speed” has 3 decimal places (0,000). If I apply 0,225 to the track, It’s still longer the the original track, but if I use 0,226, it shrinks too much. 
I think that’s all we got. Also, you’re guessing what they did between the two versions, right?
Cancellation is tricky. There’s pages and pages if reasons for it to fail. If anybody used MP3 or any other compressed sound anywhere in the process, you’re dead. If they took an analog step in production, that’s the end of the world.
I can think of no good production reason for one to be longer than the other. So Something Happened.
Koz
Change the “New length” box in “Change Speed” to hh:mm:ss + samples (click the downwards-pointing triangle) then set the speed change according to the exact length you need. But you should not assume this will necessarily do what you want it to.
Gale
If you apply the corrections either side of perfect as you have, I would expect at least the first quarter of the song to be approaching success or show signs you’re on the right track.
Cancellation can be affected by both phase (timing) and volume. If whomever made the second track decided to change the volume ever so slightly, you will have to adjust for that, too.
If your cancellation gives you Space-Ship-Mars sounds, then it’s a sure bet somebody slipped a compression step in the middle. Most compression technologies are more or less invisible until you start to play games as we are doing. It’s not good to look under the hood of compressed music. If that happens, it’s the end of the world.
Koz
This is another posting who has a cousin problem.
https://forum.audacityteam.org/t/need-to-fix-reverb-in-recorded-interview/45058/1
She has a valuable interview with a Skype echo on the voices. It says here, you should be able to use phase inversion, timing and volume to cancel or at least suppress the echo.
It’s been days. No word.
Koz
Thank you all for your replay.
Yes, i’m just guessing. I thought maybe they did that on porpose to prevent people to do exactly what I’m trying to do.
I can’t find this triangle you said. You mean going to “effects ->change speed”? I only get “percentage change” and “vinyl rpm”.
The closest thing I got from getting the voice is exactly that! When I align it and use “invert” and then play it, I get a sound like a “senoid”, as if the voice is untouched but the volume of the instruments gets low than quickly goes back high, then low and high and so. But I guess you already gave me the reason: I’m working with mp3 
I’ll re-rip it in flac and give another try.
See this image:
If you don’t see the length boxes then you have an old version of Audacity and you should upgrade: https://www.fosshub.com/Audacity.html/. We always ask you to say what Audacity version you are using (see the pink panel at the top of the page).
Gale
But I guess you already gave me the reason: I’m working with mp3
You can stop now. You can’t do phase cancellation from MP3 and once you make one, the damage is permanent. You can’t convert an MP3 to anything else. That’s the reason we recommend very strongly not to use MP3 anywhere in a production.
Koz
If you’re ripping Audio CDs, they should arrive as WAV, 44100, 16-bit Stereo. That’s the CD native format. Leave them that way.
If your ripper insists on making MP3, get a new ripper.
Koz