hi there, i talked to a guy in mirc today and he gave me some reasons not to use VBR (not just MP3 but OGG too), i found it funny but i wanted to know if there was any truth in his words. here it is:

[21:52] <ssjkakaroto> say, why dont you guys use vbr mp3s instead of 320 cbr?
[21:52] <someuser> ssjkakaroto: VBR is really bad if you listen to instrumental music
[21:53] <ssjkakaroto> really? even with --ape?
[21:54] <someuser> ssjkakaroto: yes even with ape , 320 CBR is better when listening to instrumental music
[21:55] <ssjkakaroto> hmmm, so why not switch to ogg or mpc??
[21:56] <someuser> ssjkakaroto: ogg is VBR too and mpc.... not many people know mpc
[21:57] <anotheruser> don't like vbr music
[21:57] <ssjkakaroto> someuser: i'm pretty sure ogg sound better than a 320 mp3 with q8 or q9
[21:57] <anotheruser> gives me the creeps when i play in my favorite player, sonique and i see the bitrates constantly changing
[21:58] <someuser> ssjkakaroto: ogg is VBR.... so it is not perfect... VBR always loses quality... you just don't recognize it if you don'yt listen to instrumental music with good speakers
[21:59] <ssjkakaroto> but a 320 cbr doesnt mean youll have a better quality
[21:59] <someuser> ssjkakaroto: at least it is constant quality...
[22:00] <ssjkakaroto> thats my point, the constant quality doesnt mean better quality

someuser left after this
so was he right?

VBR bad for quality? Sounds like nonsense to me...
Then again I'm not an expert...

--alt-preset insane = 320 CBR: all frames 320kbps
VBR, e.g. --alt-preset whatever else: 320kbps and lower frames, so quality is equal or worse

[EDIT] I'm quite tired ATM and will go to bed in a few minutes, so my thoughts may be not as clear as usual, but ... What seperates trolling from talking about trollish statements? [/EDIT]

constant bitrate != constant quality

I suspect them to evaluate quality with a frequency graph. With mp3 VBR or with mpc, lowpass is not constant, so they conclude that quality is good only on few parts of the file. A graphical representation of a vbr file looks like a desolate landscape ; CBR gives the feeling of an 'intact' file.

I often read such arguments on french forum. Not easy to convince people, when they don't trust their ears.

Basically I think he doesn't understand the techniques used in lossy encoders, just thinks of quality based purely on bitrate and nothing else.

"VBR is really bad if you listen to instrumental music"

"gives me the creeps when i play in my favorite player, sonique and i see the bitrates constantly changing"

"ogg is VBR.... so it is not perfect... VBR always loses quality... you just don't recognize it if you don'yt listen to instrumental music with good speakers"

"at least it is constant quality..."

FLAC and Monkey's audio are VBR too.

Not everyone wants to use the highest bitrate the codec has to offer. People want to use 192k, or so, and have it be lowered or upped when needed. If they want 320k MP3s, of COURSE it'll sound better than VBR. If you want to make a 192k CBR MP3, of course VBR will sound better.

Back with the old Xing encoders, VBR might have honestly been worse for instrumental music. That's back when I learned to distrust VBR, after hearing total crap come out of Xing. It wasn't until somewhat recently that I found out VBR didn't suck anymore.

APE is VBR? Please. How many times do people need to hear the word "lossless" to understand that there is no loss?

Yes, APE [as in Monkey's Audio] is variable bit-rate, only lossless. VBR is not restricted to the lossy domain

Cheers,
CD

Edit: Changed quote date and time to ISO 8601 format.

Clarification: APE is a Monkey's Audio file. --ape (which is what the original poster was talking about) is an abbreviation for --alt-preset extreme.


A troll won't listen to reason, but someone talking about trollish statements is asking for reasons. Of course, the reasons can't be used against the invincible troll, but I don't think this "someuser" is a troll. There are a few cases where Lame's VBR psymodel messes up string instruments (or pure tones), but CBR doesn't. Guruboolez's erhu sample is one example I think, and KikeG also submitted such a sample. In these cases, CBR 320 (--api) sounds better, or even CBR 128. But there are so many pre-echo samples for which both --aps and --api sound non-transparent, I don't think these couple samples should be reason to use --api over --aps. If someone cares that much about quality, then they shouldn't be using MP3.

Of course, Lame's psymodel isn't a factor with Ogg Vorbis. Someuser's broad dismissal of Vorbis is silly, and his dismissal of VBR in general is simply fallicious. Maybe a bit of education would straighten him out, or maybe he's just stubborn, in which case you should just hax0r him and make sure he doesn't spread any more BS.

Lies! Nonetheless, I imagine that its VBR doesn't make instrumental music sound worse...

However, the poster did seem to mention "--ape," not "APE." My mistake.

Does anyone know of any good links that go in-depth about how VBR works? I haven't found much...

I'm not sure where to tell you to turn to on that last request. However, I believe the HA irc channel should have people there that can help. I agree that dismissing VBR as a matter of course is a very bad idea. The concept behind VBR is that the encoder is free to use only as many bits as it percieves is needed to encode the audio with a minimum of errors. In this way, you can encode at (theoretically) the highest quality possible for difficult passages, but you encode at lower bitrates to save data on less complex passages, resulting in an averaged, smaller size. Of course, different encoders do this differently, and to differing degrees of quality. The way the "troll" describes it, VBR is bad simply because it doesn't stay at full-throughput for the entire song, and therefore is throwing out needed audio. This is only half-truth, as much of the audio that would normally be removed cannot be percieved by your average listener, and therefore isn't needed. I think that there should be someone in the Hydrogenaudio IRC channel that can help you out with further questions on VBR. Good luck, there are a lot of people out there who either misinform others, or have been misinformed themselves, and will lead you to answers that are far from the truth.

The highest constant bitrate will always, in theory (ie, with no buggy code or anything), be higher quality then any VBR. Of course, it just wastes bits like mad. I think MPC standard hits close 500 kbps on some frames. That just means that to get better quality then standard with a constant bitrate you would need to use a CBR of at least 500. That's why MPC can handle some tracks that even ATRAC (with a 280 CBR) trips up on. If he doesn't want to use MPC or AAC, then yes, CBR 320 (done well, like with --api) will give him the best quality.

Yes, he is right, although VBR can come close or could be better with another format... possibly. The bottom line is that CBR 320 MP3 IS better than --alt preset standard or even extreme

Sure VBR makes more since (better quality for file size), you cannot beat 320 CBR MP3

I'm pretty sure nobody will argue the fact that 320kbps CBR is capable of better quality than VBR MP3s. The point is that this person thinks there is something inherently wrong with VBR; MP3, MPC, OGG, it doesn't matter, if it's using VBR it'll fail with "instrumental music" where apparently 320kbps CBR MP3s don't.

Think of all the poor kittens...

A note about APE (as in Monkey's audio). It uses a variable amount of bits to encode a unit of time, so it's strictly speaking also a VBR codec. All lossless codecs I know are.

It guarantees constant (perfect) quality by using a variable amount of bits. It wouldn't be possible to guarantee perfect quality if it was CBR.

That's why VBR MP3 is so bad! as Sonique is evil, can't even handle VBR MP3 properly............think that guy should use Winamp or fb2k or MJ!

Well, indeed it wouldn't, unless the bitrate chosen for CBR were 1411.2 kbps

thx for all the replies, was i right saying that OGG q8 or q9 (maybe lower) would give him better quality than MP3 320kbps?

I think, not necessarily. API may be better than Vorbis at any quality setting.

Excuse me, VBR is worse quality (or better quality) compared to WHAT?

If we compare VBR file who's bitrate is variating between, let's say 160 and 320 kbps to 320 CBR, ofcourse 320 CBR would have BETTER (but not optimal!) quality.

But if we compare the same VBR file to 160 CBR, apparently VBR file would have much better quality (with lower compression ratio)

A VBR file with an ABR (average bitrate) let's say 192 kbps, when compared to 192 CBR generally would have similar or better quality, since the bitrate is distibuted much wiser over the whole file, trying to keep higher bitrates for more complex sounds.

The purpose of lossy encoding (like MP3) is to achieve MAXIMUM COMPRESSION ratio, maintaining satisfactory quality.
The purpose of lossles compression is to PRESERVE the original (BEST!) quality, trying to achieve better compression ratios.

So, if one is hunting for quality, he should consider lossless.
Lossy is never considered to be the "best quality", rather it is "optimal" quality (compared to compression ratios )
So, if we choose lossy (MP3), we are hunting for optimal, not maximum quality.
VBR files, achieving optimal (or close to optimal) compression ratio AND quality factor, apparenly are MUCH BETTER THAN CBR WITH _THE_SAME_BITRATE_.

Right?

So high bitrate ogg files are a no-no?

Is this also true for gt3?

the trick here is to trust your ears, and not what the graph shows or the bitrate indicator in winamp or sonique displays. and if sonique can't decode vbr properly, well then thats a problem with sonique, not vbr mp3s. in theory, --alt-preset insane should produce better results because it encodes everything at 320 kbits, even passages that don't really require that many bits. the issue here is whether this individual will even notice the difference between, say, --alt-preset extreme and --alt-preset insane. i doubt he will notice the difference

Lossless or MPC looks like a lot better decision.



Less so, but still.

i didn't like vbr either. the first time i listened to the file i thought there was something wong with it because the bitrate was going crazy in winamp. so i hated it for that reason, ignorance. vbr is all i use now.

garf, is there a sample that one can abx the 320 cbr mp3 from the q8 ogg?

I wouldn't use OGG for anything higher then 100kbps. It's hands-down the best for that range, but as soon as you go higher it's beaten by almost every other format there is. --api will give you much better quality then OGG -q9.

I know someone who can abx ogg -qx every time, but get's exactly 50% with --aps.

PS It ain't me. My ears suck.

While Vorbis is superior to LAME MP3 below 100kbps, I wouldn't say it's hands-down the best in this range. The majority believes that Vorbis is truly exceptional at lower bitrates but I think too many people discount other formats. Even the scoffed MP3Pro can give Vorbis a run for its money. The community has a bias for Vorbis because it is an open-source project and this bias has a tendency to distort reality. May I direct everyone's attention to ff123's 64kbps multi-format test performed this past summer.

I would use Vorbis at lower bitrates because it is free and generally performs very well but it is false to believe that Vorbis is the only sensible choice. I think that on an objective-oriented forum like HA we need to avoid absolute generalizations unless we have extensive objective testing data to support them. Yes, Vorbis is a very competitive format but it is not the best. Nothing is the best at this point.

I remember using napster beta 4 and downloading a song called blue skied and clear. The time displayed in the built in media player went crazy. I deleted the file because I thought it was corrupted. My have I been enlightened since then.

Calx has hit the nail firmly on the head up there, it's simple ignorance on the part of the person the original poster quoted. He wasn't even trying to ask whether a VBR file was better or worse than a CBR 320. It was obvious from his statements that he hadn't got a clue what he was talking about; this thread has involved some completely unneccesary arguments.

The tell tale line from the IRC guy who was quoted was

"gives me the creeps when i play in my favorite player, sonique and i see the bitrates constantly changing"

Read it again, what does he say? He says "gives me the creeps". Not "It sounds worse" or anything similar, he says that it gives him the creeps. Well, ghosts and monsters used to give me the creeps when I was a child and didn't know any better, but at least I've now learned that there's no reason to fear them... just as there is no reason to fear VBR.

Another line from that person really annoys me too, it's something that's said a lot amongst ignorant people:

"CBR gives consistent quality, VBR doesn't."

Again, this clearly demonstrates how little the person knows; exactly the opposite is true. With VBR you are specifying a consistent quality for the whole piece, that's why the bitrate will change as the complexity of the music does. With CBR you are just specifying a constant bitrate, which has nothing to do whatsoever with consistency of quality; in a CBR 128 file, the very quiet or "simple" parts of the music might sound transparent, but when the tune gets going properly it'll sound terrible. That's not consistent quality.

What annoys me the most is that you don't even have to know anything about MP3 encoding to realise that, it's common sense. Music isn't a constant tone, it is therefore illogical to use a constant bitrate with it.

torok do you know what sample does this person use? and according to your sentence this person would be able to abx even on OGG q10 but would fail +-70% if he used --api??

I'm encoding my CD's with Vorbis q 6 now, which results in a bitrate of more or less 192 kbps. Would MP3 (properly done) be better at this bitrate?

BTW (OT), Garf, are you Flemish or Walloon?

Well, psychoacoustic compression also has a nasty tendency to do this. What we hear is a distortion of the source performance.

Completely unrelated, but for a few of us, distortion of reality is all we have. If we think it sounds great, who are we to argue against the benefits of the placebo effect? Tens, maybe hundreds of thousands of geeks fanatically convinced that Vorbis sounds better is a great way to bring less-critical digital music fans onto the Vorbis bandwagon. When Vorbis reaches critical mass, we benefit by cheaper access to Vorbis dedicated hardware, software tools, and players. Corporations benefit from cheaper access to a great compression suite. Popularity draws developers, and everyone benefits from improvements to the Vorbis code base. And the cycle keeps going.

To me, it seems like a win-win situation. All started by zealots who thought they heard a difference and were loud enough about it. Lets face it, average Joe isn't going to hear much of a difference, nor does he care that other formats are patented. He may care about lower-bitrate quality if he uses a portable, but how many people that own a portable and know/care about how suck every last bit of performance out of it don't already participate in forums such as this one? On the flip-side, everybody knows a geek and trusts his opinion on technological matters.

<edit>Grammatical clarification</edit>

I wouldn't say better. I think both formats with best encoders are close at this bitrate. Vorbis can be better, but it's not there yet.

I'm Flemish.

I know Santa Claus.

Such general statemets are nonsense.

Sure, I suspect most of the problem samples submitted in the GT3 thread would fit that criteria. I don't have any where the reverse is true since I don't tune LAME

That's all fine and dandy, but the goal of this forum is to provide objective technical information. I won't deny that placebo effects are important all over - but we aren't here to promote those.

You perfectly reinforced mithrandir's comment about distorting reality.

If you consider the range as such, it's a reasonable statement (of course I'm biased). MP3Pro is as good at the low end and MP3/AAC are compatitive at the upper end, but I don't think there is any codec that performs as well over the entire range.

How is that general?

I was running the test and it was Coldplay - Yellow, from about 00:30 to 00:50. I was using the bitrate settings and have forgotten the corrosponding quality, but I started at about 100 kbps and kept doing tests up to about 210 kbps, I believe, before I just got generally annoyed and quit. Next time she's over I'll do it again and post the ABX log.

So, I didn't try all the way to -q10, so -qx was a bit misleading. Sorry.

Garf totally called you on your general post, dude. But thanks for clarifying your statement (in the post above this one).

BTW I'm glad you edited your post; the pre-edited version was pretty nasty.

Luckily we have SometimesWarrior to correct us all.

Hahahaha...

Yea. One of those things where you type it up, and then you start thinking about it and realise that the other guy was a lot more right then you origonally thought.

Somuser is the kind of people that make mu stomach hurt and make me feeel like writting giants articles about the truth.

but no matter how many times you try to educate someone. a new no-brainer starts opening. his mouth telling stuff he has absolutly NO knowledge about :-(

wish i had a weapon that remove alle people telling things they have no clue about like they know everything about it.
i just goddamn hate it..... its making it hard to spread the truth.

In my original post, I gave thanks to everyone for pretending not to notice all my nonsensical posts. But I kept previewing and editing my post until it didn't make sense anymore, and then I was late to class, so I foolishly submitted it...

I totally got called.

...and by lots of people asking what's the best cmdline for 192k CBR vorbis, IYKWIM.
Good thing is that oggenc doesn't offer as many switches as lame to mess around with, IMHO.