I want to compare different WAV files containing the same song, these WAV files come from different CDs from different record companies and the WAV files are almost, but not exactly, the same filesize.
If I use WinDiff from the Microsoft Support Tools that sometimes works, then I see that both files are the same except the beginning and the end or that they are totally different. Only the problem is that WinDiff hangs on a lot of files, so that it doesn't show results.

So now I'm looking for either a better working version of WinDiff or another compare tool with support for binary files, or another way to compare WAV files with each other.

Using some HASH-utility (no not for in your pipe) like "MD5Summer" http://www.md5summer.org/download.html will tell you that the files are the same (bit for bit), it won't tell you though if there's 1 bit different (or that it's actually a different song ftm) when they are different. I don't know what your objective is, this is one approach...

Thank you for your reply, but that doesn't work since the WAV files I want to compare are not even the same filesize. I know the files are not exactly the same, but I want to know if the difference between the files is 1% or 99%.

I think you're using the wrong approach. You're trying to see how different two audio files are by looking at the binary data. What you should do is load them in an audio editor like Adobe Audition and compare them in spectral and waveform view, and probably inverse-mixpasting both tracks (what's left will be the difference signal). As you know, half a second more silence before the start of one track can lead to 99% difference from then on, even when the actual audio is identical...

Thank you for your reply.
Where can I find how does inverse-mixpasting works, because when I search for it I get no results.

It's not that complicated. Load into Audition (or any other decent audio editor) zoom all the way into the dots the represent each bit and place the cursor on both waveforms at a point where they are equal. Then you select one of the waveforms to the end, copy it, open the other waveform, select edit and paste (mix) - In the dialog you can mark the channels to be inserted inverted, click them both and you are ready to go.

If they are equal, the remaining audio should be silence. My guess is however, that you won't get silence. If the sources are different, there might have been used different compression/limiting processors, but it's still fun to try

I think he wants to compare if the files are digitally similar, excluding digital silence at the start/end. WinDiff is a file comparison tool that searches for parts of the two files that are the same, even if they aren't in the same position. The problem with using the audio editor method, while probably more useful as it allows for small differences in the digital data, is that it will be difficult to 'synchronise' the two files, to account for the extra at the start/end.

I agree that Audition's mix/paste is not an option here, it's difficult to figure out the position to paste and above all this option is buggy (it will report differences sometimes even when there are none).
Using something like windiff is ok, you could try other tools like beyondcompare or examdiff pro (both have trial versions and hex support).

Ah i see, so he primarily wants to exclude the silence from the equasion. Then use CRC32 (see attachment cause their homepage is down), it has a special routine for WAV files that excludes silence. Press CTRL-L to load a file.

ExactAudioCopy has a WAV Compare component (ctrl+w). It take drive offset into account and shows the difference in a timetable list.

Not really. It should be fairly easy to line up both tracks by sliding one of the tracks to some sort of reference point on the other track at the sample level. Even if the head and tail are entirely different (if one of the songs has a non-silent gap, while the other does not, for instance), an identifiable transient can be used to line up the two tracks. Invert the phase on one of the tracks, hit play and watch the meters, or bounce it down and eye the spectrogram. It'd be fairly unusual that one of the songs would have additional measures or anything, so pending anything weird, this should work just fine.

Re-aligning regions in this way is a fairly common process in production, really. You might call it "eye sync"

I think you need to mention that you have to switch over to Multitrack View first. It should be really simple once you have one waveform over the other to see if the disc is from the same mastering. I actually use this in conjunction with mix-paste rather than playing back the difference (I've seen tracks with real differences that weren't large enough to trigger the meter).

menno is right, full-scale samples will not always subtract out to zero.

.
Might this work for you?:

http://theaudiocritic.com/blog/index.php?o...35&blogId=1

Andrew D.
www.cdnav.com
.

That sounds exactly like what CITray suggested above...

Thank you all for your replies.

I've used Beyond Compare on the files WinDiff hangs on and Beyond Compare works fine there.

CiTay's way of inverse-mixpasting also worked nicely.
I have also tried that with the normal version and the instrumental/karaoke version of the same song, which resulted in something like an acapella (almost only the singing) version, but you can still hear the instruments a little bit. Is it possible to filter all instruments out, so you really only hear the singing?