Checking for "fake" FLAC files


Posted this on IRC but I think it’s a slow day.

Basically I’ve found a FLAC online and oddly a comment I’ve seen says that this particular FLAC file is “fake” (their terminology, not mine!)

Additionally claiming that Audacity can somehow tell you if a FLAC is “fake” or not. They go on to say (and yes this is their spelling)… “ANALYSE SPECTRALE”. I get the impression perhaps English isn’t their first language.

Also they say (again… quoting)… “AUDACITY/ FAKE FLAC 17000 Hz !!! A TRUE (flac) 21000 Hz !”

So I’m thinking possibly they mean to use the menu option Analyze | Plot Spectrum ?

If that is what they mean, then when the dialog box appears and I hover over the graph with my mouse, the ‘Cursor’ value goes as high as 21479 Hz… which maybe means not “fake” as they claim?

I think possibly they were trying to claim maybe it was made lossy and then converted to a FLAC after, which of course wouldn’t re-make it lossless :slight_smile:

Guesses on a postcard… or better yet… right here :smiley:

p.s. Audacity 2.2.1 64-bit from main repositories of Ubuntu Mate 18.04



If you mean #audacity on IRC, I don’t think anyone uses that any more. If you saw it advertised in our documentation, please tell us where, so that we can remove it. The support channel for Audacity is here (this forum).

I’d guess that’s what they mean.

Yeah on freenode, however I just guessed the channel name (often popular projects have a presence on freenode). However I believe that freenodes policies are such that channels are only meant to use a single # if they’re endorsed by the same named project. Basically unofficial channels are meant to use a double ## to indicate this. So if it is unofficial you can always take it up with freenode, who would likely reserve the ‘audacity’ namespace (even if you don’t use it yourself) so no one else can (unless they use a double ## in the channel name, which they’d likely move to).

Is there a way with Audacity to know how true or false this claim might be?

Probably not.

They appear to be basing their claim on the fact that the audio bandwidth only goes up to 17000 Hz, whereas FLAC files are capable of reproducing frequencies over 21000 Hz. However, there are many reasons that the audio bandwidth could be limited to 17000 Hz or less.

OK so in my original post I mentioned that I’d already opened the ‘Frequency Analysis’ dialog… by selecting the whole waveform and going to the menu option “Analyse” | “Plot Spectrum…”

If I’m seeing any ‘Cursor:’ or ‘Peak:’ value on that dialog box (as I move the mouse cursor over the graph) that’s above 17000 Hz … then that is… at least ruling out the verdict that the original “commenter” was trying to make?

Have I got that right?

As we’re not even clear as to what the commenter was trying to say, I don’t want to speculate.
A lot of people make dubious claims regarding sound quality, but at the end of the day the only thing that concerns me is whether I enjoy listening to the music. :wink:

There are MP3 encoding options that allow you to keep the full spectrum* and there are other lossy formats that keep the whole spectrum. Or, someone can take a “regular” MP3 and use an harmonic exciter effect (or other effect) to “regenerate” high frequency content and then save-as MP3. It’s easy to “fake” if you really try but you’re not going to restore the sound quality.

The problem is - When you hear a compression artifact it’s not usually the loss of high frequencies that you hear. Looking at spectrum usually doesn’t tell you much about the sound quality, or at least it doesn’t tell the whole story.

The best way to identify a quality problem is to compare the unknown file to the original in an ABX test. Of course that’s not possible if you don’t have the original, and often a good quality MP3 can sound identical to the original (even if the spectrum looks different).

There was recently a post on another forum about “fake” 24-bit FLAC files and some tool that’s supposed to tell you the “true-original” format… It turns-out that 24-bit (originally from a CD) is easy to fake too. But, it also turns-out that it doesn’t matter because 16-bits is better than human hearing and you can’t hear the difference between a 24-bit original and a copy downsampled to 16-bits (in a proper blind ABX test).


  • The LAME defaults usually give the best sound quality but not necessarily the “best looking” spectrum. MP3 is lossy and it’s going to throw-away some information. If you force it to keep the higher frequencies it’s probably going to throw-away something more important! Mostly, it tries to throw-away sounds that are masked (drowned-out) by other sounds and in many cases it does a “perfect” job and you can’t hear a difference in a proper-scientific, blind, ABX listening test. That requires some complex analysis, but it turns-out that even if you can hear up-to 20kHz or higher in a hearing test your hearing is not very sensitive at those frequencies and those highest frequency sounds in the context of regular music are weak and almost always masked.

use spek then go on Google see what an mp3 looks like and a flac and check your own file against those. It will be obvious to you

Like I said, it’s easy to fake-out a spectrum analyzer like spek if you try. Or there are LAME settings that can give you a better looking spectrum (with the possibility of worse actual sound),