Lossy audio encoding aka let's pretend we have golden ears

Hello Sup Forums - I require you to perform a simple ABX test. Use whichever equipment you have: contrary to popular belief, a shitty equipment may be affected more by a shitty mp3 than real monitor headphones.

About ABX tests:
>en.wikipedia.org/wiki/ABX_test
>wiki.hydrogenaud.io/index.php?title=ABX

Cross-platform software:
>lacinato.com/cm/software/othersoft/abx
(Windows goys may use foobar2000.org/components/view/foo_abx )

I'm shipping three FLACs.
One of these is the source file.
One is derived from an MP3, obtained from the source. Relevant switches:
-m j -V 1 -q 0 --lowpass 22.1 -b 128 -Y -F
One is derived from an M4A, obtained from the same source. Relevant switches:
--afterburner 1 --bandwidth 18000 --profile 2 --bitrate 224k

To minimize cheating I've chopped off higher frequencies (AFTER encoding). Please don't cheat.
Post results.

my.mixtape.moe/bwoesf.zip

(tripcode will be used only once to post solution and then released)

Other urls found in this thread:

wiki.hydrogenaud.io/index.php?title=Opus#Music_encoding_quality
my.mixtape.moe/yytzit.zip
twitter.com/NSFWRedditGif

welp bumping just once with doggo
go, doggo

personally I believe I'm able to tell the difference between the original and the other two, even if I can't distinguish mp3 from m4a

forgot doggo

Kill yourself reddit.

But what is the question you want people to answer?

1) perform ABX test
2) post results
3) ???
as simple as that I guess

>nigger music
Yeah no thanks.

Why did you alter the files? What music is this? You probably can't tell if it was mastered incorrectly.

>MP3
Use OPUS

in case you don't know, it's easy to spot a lossy codec just by looking at a spectrogram. frequencies above 16~18MHz are anyway near to be irrelevant as far as the overall quality of the audio is concerned
the music is irrelevant

The music is relevant memeboi.

opus is a big meme, nothing else. Above 128kbps does not have any advantage over aac or even mp3. Furthermore. there's zero support on anything that ain't a tinkered sansa clip.

no, it's not. no, it's not haspichord. rather than asking just listen to it.

Why are you fucking with the dynamic range and filtering out any compression artifacts?

no, I'm not using any dynamic range compression.

ive gotten the difference between flac/wav and mp3 like 7/10 times but you have to really be listening and almost guessing

>7/10
>almost guessing
>almost

Average audiophile right here.

nice, which one is the original flac/wac in your opinion?
"xah3Eez7.flac", "eiB1uY4e.flac" or "IeVe1ja9.flac" ?

lossless audio is a meme

Here is the spectrogram difference between
eiB1uY4e.flac & IeVe1ja9.flac
eiB1uY4e.flac & xah3Eez7.flac
IeVe1ja9.flac & xah3Eez7.flac

$ gm compare -verbose -metric MAE eiB1uY4e.flac.png IeVe1ja9.flac.png
Image Difference (MeanAbsoluteError):
Normalized Absolute
============ ==========
Red: 0.0127551970 835.9
Green: 0.0016683895 109.3
Blue: 0.0080622311 528.4
Total: 0.0074952725 491.2
$ gm compare -verbose -metric MAE eiB1uY4e.flac.png xah3Eez7.flac.png
Image Difference (MeanAbsoluteError):
Normalized Absolute
============ ==========
Red: 0.0154077765 1009.7
Green: 0.0018802399 123.2
Blue: 0.0095357415 624.9
Total: 0.0089412526 586.0
$ gm compare -verbose -metric MAE IeVe1ja9.flac.png xah3Eez7.flac.png 3.png
Image Difference (MeanAbsoluteError):
Normalized Absolute
============ ==========
Red: 0.0085023415 557.2
Green: 0.0010402342 68.2
Blue: 0.0054135637 354.8
Total: 0.0049853798 326.7

I don't believe it's possible to hear a fucking difference without any other references. The least difference (in absolute terms) seems to occur between IeVe1ja9.flac and xah3Eez7.flac; so, either these are the two lossy (and for some reasons they behave similarly) or one is the original and the other one is the most preferable lossy. Usually lossy codecs do not behave similarly

b-but muh 24bits 96000Hz flac

>3.png
oops, bad copy paste, it wasn't there obviously

My guess:

eiB1uY4e = mp3
IeVe1ja9 = m4a
xah3Eez7 = flac

>three options
>accuracy: 1/3
>Confidence that your results are better than chance: more than 1/3
jokes on you, I was totally clueless tbqh

>big meme
Opus easily beats MP3 - Hell, even 256 kbps Opus is way better than MP3 320 CBR :D

It has more minor song details, more than MP3 ever could handle at maximum quality!

Just compare a 320 CBR MP3 and a 256 VBR Opus and watch out for halls and echo effects - Opus comes very close to a source lossless file. If you render a 512 kbps Opus out of a source file, you literally can hear no difference at all anymore between these two.

Tried it myself countless of times.

Here's my guess
eiB1uY4e = mp3
IeVe1ja9 = flac
xah3Eez7 = m4a
Basically haven't got a clue.
I have a feeling this is just a big meme and that they're either all lossless or all mp3 or some shit like that.

No, it's not. Opus is extremely efficient at lower bitrates (64/96 kbps), an area where an MP3 would play like horseshit.
At 320 CBR or even V0, or against an AAC > 128 kbps, Opus is not competitive *at all*. Actually, it's advised against.
You don't have to believe me, believe wiki.hydrogenaud.io/index.php?title=Opus#Music_encoding_quality
at something as low as 160 kbps old codecs like MP3 are competitive with Opus. No listening tests have ever been conducted above 128 for Opus. Moreover, MP3 is a free as in freedom and as in beer codec support even on toasters. VBR gives you an advantage in size even over AAC/M4A if used cleverly, and the multi-decennial codec LAME with its Y switch makes wonders.
If you're worried about more minor details you won't be able to listen, AAC is the only sensible choice
Have you ever conducted an ABX on it or do you hear it because you know which is the opus and which is the MP3?

close
close
We have a winner
"xah3Eez7.flac" == M4A
"eiB1uY4e.flac" == MP3
"IeVe1ja9.flac" == ORIGINAL

Since the Opus meme has been discussed, here's a track with wide imaging and different details. One is the original, one has been transcoded to Opus, one has been transcoded to MP3. To give Opus even more advantage, MP3 is V0.
For Opus, the highest suggested settings in opusenc has been picked (VBR is the suggested default). Relevant switches:
--bitrate 256
For MP3,
-m j -V 0 -q 0 --lowpass 22.1 -b 128 -Y -F
The source file was in [real] 24bits. High frequencies have been cut _after_ encoding to prevent cheating* & no filters nor compressions have been applied.

*to be honest high frequencies would have been probably the only reasonable way to tell these files apart, if you have really good ears

my.mixtape.moe/yytzit.zip

(keeping trips for another post)

>no more replies
k, giving out the second solution and releasing trip + polite sage
ihaviaC2.flac == ORIGINAL
wughoh9A.flac == OPUS
taChouc5.flac == MP3

tripcode: #eef8eQua

Rest in Pieces, Sweet Prince