That's good considering that 22050 is the nyquist limit itself for 44.1k systems. I am not sure I can really hear above 18 or 19 anyways.

 

I've sorted through thousands of duplicate flacs, how I do it:
1) use foobar2000s file validity checker to make sure the flacs are intact and the file size is not different due to CRC errors
2) use your ears to choose one to keep

 

Nulltest is a good idea as people here found out, but use an analyzer and see how many dBs of volume the Delta signal has. That gives you a good approximation on how much actually got changed and where (on the frequency spectrum).

You have the same problem with Nyquist filters in your DAW/Plugins - sure, lowpassing at 20khz with a very steep filter wouldn't have any audible result, but those LP filters usually go down to our hearing range ~16-18khz

 

As good as the 'null' test is, I don't need to do any of that. I'd play it in the car when the missus in the passenger seat. If she moans that it's too loud? I'd go with the other copy.

 

Just for the books, if you're doing a nil-test you can only find out whether there's a difference or not, not which one sounds better.
Other than that, the others are right, two flac conversions of the same file with different compression still sound the same.
So to find out which one is better, make a dynamic and frequency comparison and a blind test.

 

That's exactly the reason I stopped using FL's Free Filter, there was a very noticeable change in the 16-18k region even if the filter was fully open. The filters in Love Philter are better (might be the same as well, but they sound good at defaults). I use kilohearts' filter now, it has 20k limit as well but it doesn't affect the 16-18k as much, not noticeably at all.

 

LAME has no lowpass for V0 VBR btw

 

I did a null test on Reaper and there was just silence in the master meter.

I still haven't checked the answers that went into the area of a FLAC may be a transcoded lossy file, which is also very much of interest to me. Once I do I will get back to this thread.

Thanks everyone!

 

Words of true wisdom. Why this isn't already marked as "best answer"?
We all love music, but there're lines we can't cross. For chrissake guys, get your priorities straight!!

Nice, I'm not the only one who remembers VBR! In a decent world all lossy formats should be VBR, just like in video compression.
Then there's the sad real world. Hug for you

 

It's honestly a legit question but I still have to wonder, if you can't hear a difference, does it matter?

 

well I'd rather delete installers for all those mediocre daws, plugins and kontakt libraries in first place

 

Haha good one. I do use no install versions. But no install versions aren't smaller depending on the program/plugin. Well, unless you use 7Zip with cowboy/ballz to the wallz compression profile.
And I "slim down" Kontakt libs. And beer bottles...

 

This thread ranks very highly on Google when looking for a tool to compare the audio content of FLAC files (which is how I found it), so I thought I'd add the solution I found even if it's an older thread. It should be quicker and simpler than nulling tests. Edit: and I noticed it's mentioned a few posts above, must have flown by my eyes… sorry!

foobar2000 has this functionality built-in. Great if you're already using it as a music player anyway, like me, but it's also a small and non-intrusive software in case you want to use it just for this.

Simply add both of your FLAC files to the playlist, select them, and from the right-click context menu choose Utilities > Verify integrity. Apart from checking the files for errors, it also calculates MD5 and CRC32 checksums of the audio streams. The checksums are calculated on the decompressed stream and discarding any metadata, etc. so if the files were differently encoded but from the same source, the results should match.

I verified this using two FLAC copies I had of a CD, different in both bitrate and metadata, but each track of the two copies resulted in identical checksums.

 

Lossless
Identical
Compression is not about quality, but other data.

 

This is excellent information and makes it trivial to compare against AccuRip databases for losslessness of the source at scale. If someone wants to rebuild this in the command line for speed and less tedium:

ffmpeg -i "input.flac" -f s16le -acodec pcm_s16le - 2>NUL | checksum

Outputs the CRC32 of the raw PCM and is identical to the checksum from foobar2000's 'Verify'.

(checksum.exe taken from libdeflate, NUL is the Windows null device)

 

Correct me if I'm wrong, but FLAC files encoded at the highest degree of quality are in fact lossless when unraveled back into an uncompressed PCM format.

 

FLAC itself is always lossless, but only supports 16, 24 and 32 bit integer formats. So if you print to 32 bit float WAV, you'd have to perform a lossy conversion to 32 bit int with another tool for FLAC to accept the file in the first place.

 

https://www.spek.cc/

 

That is why you use WavPack for 32bit float files and not FLAC. Both have very similar file sizes, so ...

 

unless your target audience is a dolphin no one will notice. life is short, find something else to obsess over. Like will the Indians ever win the World Series, or is the color blue I see, the same as the color blue you see? or why doesn't Godzilla just stay home instead of going to Tokyo to get his ass kicked on a regular basis?