fakeflac
fakeflac copied to clipboard
mevdschee
Only relies on cutoff?
was wondering if the tool only has cutoff as rely, even though this is the case, it returns 100 on files with -16k cutoff
The script only relies on cutoff to determine whether a low band filter was applied as part of a MPEG compression algorithm. There is nothing else the script does. I'm not experienced enough to read your picture nor do I know what a "-16k" cutoff is. I have no idea what compression was applied or what you wanted the script to detect. If you uncomment the "matplotlib.pyplot" line in the script you can see the frequency diagram and understand better what the script has seen and why it reported what it did.
NB: If you upload a (partial, 10 sec max) audio file I can analyze it for you and give you a better reply.
@mevdschee this is what i wanted to make sure of and now you confirmed that it has no other relies. the file (tst0_upsampled.flac) has a cut off below 16000hz and still returns 100. tst.zip
plot figure
the image attached earlier is a spectrogram of the file which shows frequency loudness over time.
added a second file tst1_upsampled.flac in the zip if you'd like to have a look at, case is different where it has dynamic cutoffs.
really appreciate it.
from this article mentioned in the README.md Detecting 'fake' FLAC files : 'The Nyquist theory says you need to have a double sampling frequency in a discrete signal processing system (e.g. 40Khz)' which makes spectrogram and plots has same concept with doubled sampling frequency for plots, 16000 (16khz) = 32000 on plot.