This is why you should always verify that your FLACs aren't lossy reencodes after you finish downloading them

This is why you should always verify that your FLACs aren't lossy reencodes after you finish downloading them.

Attached: scrot.png (699x74, 18.66K)

Other urls found in this thread:

spek.cc/
fakinthefunk.net/en/
twitter.com/SFWRedditImages

I do the same but I encode anime with MPEG-4 and claim it's h265

how to tell?

You can't. That's the point.

Attached: here's your flacs bro.png (797x264, 80.76K)

spek.cc/
fakinthefunk.net/en/
Also I get all my FLACs off Deezer so I don't have this problem

There are programs like aucdtect that checks if it's real flac. To trick the program I usually add low level pink noise at around -45-50db to mp3, convert it to flac and it shows as 100% CDDA quality. It has frequencies above 20k hertz and retards think it's lossless.

yep this was me lmao

Attached: p.jpg (199x253, 9.22K)

spectrum analysis
each lossy algorithm has their own telltale quirks

Attached: file.png (1920x974, 3.81M)

oh nice

This thread sponsored by the RIAA.

with your ears

mp3 256 is noticeable, 320 is almost the same as flac; nobody could reencode anything lower than 320 and get away with it

Good thing I dont torrent music and exclusively use spanish blog sites that offer direct 'descargar' links.

what does the logarithmic scale refer to? why are decibels negative?

>why are decibels negative?
Because it's a logarithmic scale...

>torrent album.flac
>burn cd
>play on my stereo
>sounds like shit
>spectrum analyze file.flac
>tops are flatter than Sgt. Joe Friday.
>rm -r album.flac

>mp3 256 is noticeable
Have you done a blind test? How old are you?

hash from uncompressed cd source

At 96 kbps you should just be able to tell with your ears unless the master is giga shitty, but then you have bigger problems anyway.
If you don't want to listen to the whole thing attentively, you can look at the spectrograms or use an automated program. The best I've found is auCDtect. Unfortunately, it's an old closed-source Russian program. The theory is way ahead of the available software here. I have many good papers about this question saved which have no public implementation.
I was going to start a project to write a modern replacement to auCDtect based on these papers but I don't have the time for this right now. If anyone wants to take up the challenge, I can point you to the resources.

Cool LARP, clown.

so what are they related to? what is the 0db?

>so what are they related to?
Volume of the sound.
>what is the 0db?
Whatever reference point you chose for your scale. In case of the OP's graph I'm guessing it's the maximum level possible in MP3 as all the levels are negative dB.

but sound is already measured in decibels and the scale goes up...

OP's graph probably just uses a different reference level than common sound measurements. Do you know how decibels work in general?

airpod user cope. his ab/x testing came back 60k