http://www.head-fi.org/forums/showthread.php?t=227727

What does this mean??? Does this mean that the most identical looking to the original sounds best? I posted this here cause I want a developers view of this

There have been many questions like yours...

short answer is: the sense of hearing has other preferences than the visual sense. Comparing spectrograms does not say a lot about audible quality

We see these misguided individuals from time to time. The whole point is that these visualisations demonstrate absolutely nothing about the quality of a lossy encoder. The only way to establish the relative merits of the different encoders at differing bitrates, etc., is to use your ears - they were made for hearing!! The eyes were made for seeing, just to state the obvious, and you wouldn't dream of using your ears to decide on the relative merits of the pictures produced using different digital camers would you? This guy obviously doesn't understand anything about the principles of how lossy encoders are designed to work. Ignore him.

Probably that Head-Fi is pretty worthless for anything but speakers and headphones.

The reason why Blade (nowadays an obsolete MP3 encoder) is described as "almost perfect" is the lack of a
proper psychoacoustic model. Blade (and Shine, too) just encodes without exploiting the characteristics of
human hearing.
Vorbis on the other side probably does just that, what explains the somewhat ugly frequency graph.

Isn't it amazing how many people are taken in by those pictures, even experienced head-fi member with a post
count of 1000+.
But if it gives you peace of mind, use Blade instead of lame...

hehehe I have actually abxed lame and blade before and lame won hands down. I just wanted your takes on this though cause the 224 vbr aac looked darn tempting and perfect I couldn't just ignore the post that quickly.

I almost feel like bothering to register on Head-fi to respond to that.

Any chance somebody can get a Jpeg at 2 different quality settings, somehow convert them into an audio file of some description, then ask them which picture looks better based on the sound?

Also, the main reasoning in some of those peoples arguments is that a graph is scientific, so anybody care to post this?

Actually this should be possible. There was an interesting thread on this topic quite a while back.
http://www.hydrogenaudio.org/forums/index....showtopic=14103

I feel especially sorry for that one sap over there who looked at the graphs then asked how to configure Blade to work with EAC. Yikes.

is r3mix making a comeback?

Hahahaha. Gawd even at 320, blade was bad

Don't worry about Head-Fi. There are enough people who know that spectrograph's aren't the best method to determine whether a codec is good or not.

And I might just post that picture in the thread. I had a good laugh just now.

It has already been post by someone else. I guess I won't have to any more.

you should. most current lossy codecs are almost transparent (in the sense you cannot pinpoint the compressed one from the original) at 128kbps.
This was proved by a controlled and serious listening test that used several subjects, real music and statistical techniques to reach an acceptable confidence level.
IMO this is much more scientific than the graph approach vented by the guy at head-fi. By reading that thread I got the impression he and many others do not know what they are talking about and are spreading misinformation. Sad.

...because clearly a 664x212 jpeg includes all the information present in a 6-second long 44.1kHz sampled wave file. (4233600 bits, if you're interested - about 0.5MB, ignoring the fact it's stereo).

There couldn't possibly be any details in the .wav file which aren't present in the 88.2KB jpeg.

Heck, forget mp3, let's just use jpeg for audio!

Cheers,
David.

Have you conducted any ABX tests to back this up? If so, please post the results.

http://www.hydrogenaudio.org/forums/index....st&p=149481

Sorry for OT...
But having browsed this site for a few months now, it seems that head-fi comes up with some pretty absurd content...

When ever I see them mentioned its like
head-fi = n00b (to cut a long story short)

oh come on... stop going overboard with the TOS 8. blade being bad is a known fact, has been proven such in the past and nothing has changed with it, so there's absolutely no need to post further tests just to make you happy.

Further tests?

Let's see a link or two to some tests showing that Blade 320 is bad.

This is a clear-cut violation of TOS 8. We back up our "facts" with evidence around here, drumliner. This is what sets this forum apart from those like head-fi. Let's not be hypocrites.

While you're at it, provide some evidence showing that I've gone overboard with TOS 8...

true i don't post much round here, but if you check the date i joined, you can clearly see i've been here a lot longer than you, so you can spare me the lecture about the ins/outs of this forum. go search for the proof yourself if you're not aware of it, why should others do the work to compensate your lack of info?

edit (you added the last line while i was already posting): to me calling up TOS 8 about a known fact is going overboard. stuff doesn't need to be proven over & over again for those that might have missed it the past.

Gee, you've been really helpful.

At this point I'm simply asking for a link.

whatever makes you happy. now go find me a link or two that proves i really couldn't or stop claiming it.

Why be so evasive? If you can you should.

...or are we supposed to take you and Donnus at your word?

I'll be honest, I thought this was a joke when I saw it, but obviously it wasn't intended to be!! How about we stay on topic and forget about a supposed TOS 8 violation that is of no real interest to anyone?

I really don't know what else to say there. It's as though they are deliberately evading serious critical discussion of the merits of the methodology. I'm trying to be as helpful as I can, but maybe the demand for such discussion is simply not there; I don't know.

I appreciated your efforts!

Well that's something then

Wow, I'm member #17 and I only have 37 posts

This was the most funny so far: http://www.head-fi.org/forums/showpost.php...p;postcount=158
The guy/girl is so stupid that it hurts!!
He/she refuses to understand the point of the discussion and keep making a fool of himself/herself.
God, I had a good laugh.

Well, I don't want to end up with us calling a bunch of people complete idiots or something. I think a lot of the problem is that few of the people in that forum are particularly familiar with what psychoacoustics and lossy encoding really are, or what the theory behind them is, and furthermore don't know very much about the methodology of testing them, or what is actually being presented in these graphs. So, it's probably not particularly apparent then that those graphs are far from sufficient in resolution (and are not even axis matched; do a difference comparison and notice that they don't line up), and that the test sample doesn't provide a comprehensive analysis. I can see artefacts in the graphs but they're difficult to make out and very difficult to identify due to such low resolution in the graphs and the inherent difficulty of intepreting a spectrograph when it's an 8 bit file (256 levels) representing a 16 bit file (65536 levels) for intensity coupled with a granularity of frequency far exceeding what would be visually perceptible given the size of the graph.

In short, the methodology involved here is outright abysmal. Even if one were to sign on to the proposition that you could dispose of the issue of codec quality by looking at spectrographs, this particular presentation and methodology is highly unlikely to produce meaningful results. I'm really not sure how to convey that; maybe I should just say what I said above, I don't know.

A lot of people fundamentally don't understand what distortion actually is, and believe that all distortion can be readily identified on a spectrogram. And a lot of people don't understand distortion artifacts in lossy encoders. So they have access to this tool they think they understand (a spectrogram), which by the way they have absolutely no idea how to set up properly, and suddenly, every audio problem is a nail...

Pretending to be knowledgable is a favorite pasttime on Head-Fi. It's not all BS though.

My favorite post


Goodness!

Those spectographs look pretty cool though

Easy enough to invalidate...
lame -b 64 --resample 44 -k
OR
oggenc2 -q -2 --advanced-encode-option lowpass_frequency=99
I can't help but wonder how that would look in their little pictures

... not to mention they used JPEG for the images instead of png

Actually ive canned blade a long time ago. I really don't see a point in abx because blade doesnt even sound hifi enough to care about. I guarantee you a 100% perfect abx result since I can even tell between wav and 320 lame 3.97 and have posted the results of that in the past. At least the differences with lame to cd was very subtle and could almost just be felt more than be heard(backed up by a 7/7 abx test i posted here before). There was a very quick cymbal hit that the mp3 just couldnt replicate. I'm pretty sure most people couldn't hear the difference though cause it was so subtle. The difference was that the wav snap startled me while the mp3 lacked a little in the attack to make me jump up (again backed up by an abx test posted here).

For most music though even lame 3.97 vbr fast standard is transparent to me and too much hassle to abx. Blade is the opposite. It's too ridiculous to abx

I don't understand why people don't understand it! It is not complicated...

The graph is so low resolution that it's easy to make a file which sounds clearly different from the source, but produces an identical looking graph.

Conversely, it's easy to encode a file where the changes are inaudible, but are clearly visible on the graph.

This proves that the graph doesn't reflect what you hear.


Now, if you really understand this stuff, you might be able to make some predictions from the graphs. You might, for example, be able to dismiss some visual differences because you understand psychoacoustics and know they'll be inaudible. You might also be interested in what the graphs reveal about the internal workings of the codecs (audible or not).

However, even the brightest and most clued up person here isn't going to be able to look at a graph which looks the same, but represents something which sounds different, and say "that will sound different"!

Cheers,
David.

So, what do you suggest? I see in that thread that for 8 or so pages people are trying to make the OP understand that his method is flawed to no avail. The guy simply doesn't care about it. He thinks he is right, period.
I don't understand why you care so much for such an 'intellectually underprivileged person' if that choice of words makes you feel better.... move along people, you cannot save someone that does not want to be saved.
That thread closing is long overdue IMO.

320kbit VBR? What an idiot...

I would suggest: make a 3D sculpture from audio, you can get a better feeling from it.

Damn, that would have made a good prank.

Wood probably feels warmer, but marmble has higher resolution. Hmmmm, what colour ?
sigh