Skip to main content

Topic: [TOS #8] Coding for Public Distribution (Read 8313 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • randar
  • [*]
  • Banned
[TOS #8] Coding for Public Distribution
ENCODING?  Releasing your encodes to the public?  Please, for the good of the community, use one of the following two command lines.  The first offers maximal portability.  The quality achieved is still very high--noticeably better than simple-stereo files encoded at constant bitrates of 192 kbps.  The second line leaves absolutely nothing to desire in the way of quality, for it squeezes every bit of juice out of LAME 3.9x while still retaining reasonable compression ratios.  Both lines are quite speedy, especially #2.

LAMELINE #1, for superior portability:
-vbr-mtrh -h -V0 -mj -b96 -Y -Z


Because file sizes are even lower than those produced by r3mix, encode times are comparable, and due to the reliance on the latest LAME series--3.9x--for improved efficiency, this line is being touted as the new r3mix (or "the r3mix killer that was" ).  Audible quality corresponds to --alt-preset fast extreme.

LAMELINE #2, for superior quality:
-vbr-old -q1 -V0 -mj --lowpass 19.6 -b112


Producing file sizes in line with those yielded by --alt-preset standard, this line amazingly manages to surpass --alt-preset extreme--even insane--in perceived sonic quality.

But SIR, why not just use the alt presets??
In the seminar I delivered today, I was quoted as saying:
"While LAME's alt presets are good, there is better, mainly because of the psychoacoustical model utilized by these presets; it is an experimental mode that has just recently been proven to have very poor pre-echo control when long blocks are subject."  For more on this, please refer to < http://www.hydrogenaudio.org/forums/showth...p?threadid=1196 >, where JohnV concludes:  "Despite Dibrom's --alt-preset insane tweaks, nspsytune's long-block "problem" still is there."  Additionally, I went on to say:  "-q1 is specially tuned for gpsycho, so where quality is #1, line #2 comes through."  To add insult to injury, nssafejoint seems to be playing a key culprit in audible distortion in a fair number of samples encoded with the alt presets; just ask "Wombat" (http://www.hydrogenaudio.org/forums/showth...hp?threadid=874).


JOINT STEREO is utilized by both lines for reasons I am sure most HA forum-goers are aware of.  There is no excuse not use VBR so it is thusly activated.  If your player does not support VBR, return it, for VBR is part of the orignal MPEG [Audio] Layer 3 specs.

Try 'em out on your personal music collections.  I think you'll be pleasantly surprised .

And remember, SAY NO to simple-stereo-192 albums encoded by the "ripping groups"!  Just because this was the way to go four years ago doesn't mean it speaks accurately for smart, efficient, high-fidelity mp3-encoding today (and it DOESN'T!).

--
Some amending humor I came across while performing a mini presentation [on the superb Lamelines above]:
<Annachie> vbr rips are crap
<Annachie> and lame is a crappy encoder
<Annachie> codec
<Annachie> you better check http://www.mp3newswire.net/stories/2000/bestsound.html
<Annachie> read about vbr and why radium codec is used by the "pros"
I asked exactly which pros he was referring to, and stated that MP3 is not used professionally:
<Annachie> some decent ripping groups
<Annachie> like eNt or JUST
<Annachie> and by no means should u use joint stereo
<Annachie> they dont have to reripp anything cuz they are certain about their releases being good
<Annachie> lame rips are always glitchy, and the bass goes high as hell
I added that FhG has long since abadoned MP3 while LAME is being continually improved to this day:
<Annachie> yet it hasnt matched radiums quality
<Annachie> so, keep on updating it
<SpinDoctr> lame needs to be continiously improved so it can catch radium one day
<SpinDoctr> it says lame iz lame (In regard to the page posted by Annachie.)
<Annachie> 160 kbps should be in Joint Stereo/MS Audio, 192 kbps should be in Stereo
--

[span style='font-size:9']Remember, we're all in this together!  With this edit I surrender.[/span]

  • Jon Ingram
  • [*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #1
Quote
Quality corresponds to --alt-preset fast extreme... Producing file sizes inline with those yielded by --alt-preset standard, this line amazingly manages to surpass --alt-preset extreme--even insane--in perceived sonic quality.


Please refer us to the results of an objective test you've carried out which confirms these statements.

  • Wombat
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #2
lol!

The samples where pre echo is a bit worse don´t justify
the absense of nspsytune.
There would be "much" more problem samples without nspsy.

When you are talking about me, what i said, read carefully.
The thread i offer the distorted samples says that is only
an isue with noise shaping. I first thought it is nssafejoint
in the end it is clear -Z beeing used with liebestod solves
the problems and sophia and birds get much better

a short test showed that no song is better with your settings
against aps -Z
And these are only samples that are "very rare" with aps,
your commandline hasn´t much potential against the alt presets!

Lameline #1 must be a joke to compete aps at ~192 kbit, isn´t it?

Believe me

http://www.hydrogenaudio.org/forums/showth...hp?threadid=874

Wombat
Is troll-adiposity coming from feederism?

  • xmixahlx
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #3
i can see the bloodbath now...

a few questions though, before you live no more...

...where are you to be giving a seminar on lame commandlines?
...who quotes themselves in their own posts?

later
mike

  • Wombat
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #4
Well to end this discussion from my point of view.

Lameline #2 has all faults we tried to tune for a while now.
Dropouts will occur like "Serious Trouble" (btw, i rarely had
heard this that badly encoded as with Lameline #1)

Block switching isn´t as efficient anymore (dibroms Tunings),
fatboy, drone...

With metal bitrates bloat.

To this list surely can be added much more stuff, but this is
enough i think.

Wombat
Is troll-adiposity coming from feederism?

  • rjamorim
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #5
ffc?
Or maybe Beatles?

LOL...
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org

[TOS #8] Coding for Public Distribution
Reply #6
This looks like a troll post to me, except it's so long... most trolling authors wouldn't spend this much time typing a post purely to start an argument.

Congratulations, Hydrogenaudio has reached critical mass

  • Wombat
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #7
mmh....

Edit after Edit!!??

J?
Is troll-adiposity coming from feederism?

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
[TOS #8] Coding for Public Distribution
Reply #8
Instead of responding to this lengthy post quote by quote, I'll just ask for specific samples where the supposed command line given is superior to --alt-preset standard.

Also, as stated by others before me here, it appears you have taken the statements of JohnV and Wombat completely out of context.  In addition, you are mistaken about the nssafejoint issue altogether.

It might surprise you to know that both of these very people were some of those who actually helped --alt-preset standard to become what it is.  As Wombat has stated, although there might be those few problem samples mentioned in the particular post you linked to, the samples which are problematic without --alt-preset standard are far greater.

Both JohnV and Wombat (and many others with great ears) have in the past performed extensive listening tests with --alt-preset standard and I'm quite sure that they would both tell you that there are no other command line options or presets currently available in LAME which will give you better quality at the same bitrate.

Anyway.. I await some sort of proof of your claims.

  • JohnV
  • [*][*][*][*][*]
  • Developer
[TOS #8] Coding for Public Distribution
Reply #9
lol, randar or should I say ffc..?

Last time on irc you wanted to use non existant switches.. I'm glad that now you have found at least switches that exist.

However, just for your info...anonymous proxies will be banned.

This guy has been banned from #project_mayhem, #vorbis and #r3mix irc-channels. I guess no need to explain more..

[edit] Except that "Lameline1" doesn't use vbr-mtrh. It uses vbr-old. Correct switch for vbr-mtrh would be --vbr-mtrh.
Juha Laaksonheimo

  • Dibrom
  • [*][*][*][*][*]
  • Administrator
[TOS #8] Coding for Public Distribution
Reply #10
randar, ffc, or whoever you are, if you wish to continue posting to this site (if you are not already banned under another alias) you will have to use a regular isp connection and not come to the site through an anonymous web proxy.

  • Trelane
  • [*][*][*][*]
  • Developer
[TOS #8] Coding for Public Distribution
Reply #11
Just where are these presentations and seminars given? I would surely love to attend!

[TOS #8] Coding for Public Distribution
Reply #12
lol determining used codecs by copyright etc flags, it's just too ridiculous

[TOS #8] Coding for Public Distribution
Reply #13
too sad that the last revision was in 2000

  • fewtch
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #14
If people *must* use one of the other "free" codecs and encode below 192kbps, I sure wish they'd choose Xing instead of Blade.  If it's a question of "bad" vs. "horrendous" there's no contest. 
Bring back dynamic range... www.loudnessrace.net

  • rjamorim
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #15
Quote
Originally posted by Benjamin Lebsanft
lol determining used codecs by copyright etc flags, it's just too ridiculous


Benni, that was one of the ideas Naoki used to create mp3guessenc. It's not ridiculous.
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org

  • rjamorim
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #16
I few minutes ago, Randar contacted me on IRC asking me to post a reply from him. I won't post it here, because I believe this issue has gone a little too far, but I will make it available at my server:

http://www.inf.ufpr.br/~rja00/randar.txt

Also, I will not make any comments or post my opinions. But rest alerted that the contents here are similar to those of his earlier post.

Regards;

Roberto.

Edit: I believe I should make this clear: Not a word of what he wrote was edited.

Edit2: The original text from randar I uploaded had some errors (paragraphs truncated). A complete one is now available.
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org

  • fewtch
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #17
"In just the past few days I have literally converted thousands..." :rofl:
Bring back dynamic range... www.loudnessrace.net

  • ff123
  • [*][*][*][*][*]
  • Developer (Donating)
[TOS #8] Coding for Public Distribution
Reply #18
As a quick and dirty test, I tried castanets, one of the samples recommended for trial in ffc's -- oops, randar's off-forum post.  I encoded using lame 3.91 aps and also randar's lameline 2 command line.

I used Ivan's waverate to have the original present as well when I listened to the samples.  Then I performed 10 waverate tests (I could have performed a direct ABX comparison, but this was in response to randar's objection that the original should always be present).  I rated ll2 worse than aps 9 out of 10 times (confidence = 99%).

Surprisingly, in chat, randar said he could not tell the difference when listening through speakers (although he says he has access to Orpheus headphones!) between the original castanets, apx, and ll2.

Then he said that the "differences so subtle and in such a short excerpt of such a small number number of clips aren't worth fussing about."

Hmm. ok.

ff123

[TOS #8] Coding for Public Distribution
Reply #19
Quote
Originally posted by rjamorim


Benni, that was one of the ideas Naoki used to create mp3guessenc. It's not ridiculous.


but you can set flags in every encoder. Just turn them on or off. In my opinion you can't say anything about the encoder because you don't know if the user changed them

  • rjamorim
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #20
Quote
Originally posted by Benjamin Lebsanft
but you can set flags in every encoder. Just turn them on or off. In my opinion you can't say anything about the encoder because you don't know if the user changed them


Yeah, but since MP3 doesn't feature any kind of header (except LAME and latest Gogo) like MPC, that says which encoder was used and, in the case of MPC, how, you must grab any kind of info you can get to try to determine the encoder. In this case, Naoki gets the default flags - Since most users just execute something like mp3enc -br 128 Music.wav, and never care about copyright or original flag, or whatever.

Regards;

Roberto.
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org

  • fewtch
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #21
OK, let's look at this one final time (I won't post on the topic after this).

Here's his claim:

"the lamelines just offer
improved efficiency, faster encode times, and better throughput."

OK first point.  Improved efficiency, what's that?  What kind of "efficiency" do these alternate command lines offer?  They sure take longer to type in than "--alt-preset standard" (and are harder to memorize), so keyboard/brainpower efficiency is not one 

Second, faster encode times.  OK, that may be, but unless running a very slow processor or encoding hundreds of files at once, what's the difference?  You only encode once, but you listen many times... and most front ends for lame allow for batch processing in the background while doing other things.  This is a minimal issue for most people.  And that doesn't even take into account that the ABR and CBR presets offer better quality than these standard command lines, and probably more speed too.

Third... better throughput.  Again, what does that mean?  Throughput??  That makes even less sense than "improved efficiency."

OK, I'm done feeding trolls and "personal command line fanatics."  We now return to our regularly scheduled programming...
Bring back dynamic range... www.loudnessrace.net

  • mpcfiend
  • [*][*][*]
[TOS #8] Coding for Public Distribution
Reply #22
Heh. Why is anybody still using MP3?

I mean, aside from the obvious portable uses. MPC is much much better for quality rips, and there is an MPC-MP3 transcoding tool available for portable users.
The sky is blue.

  • ff123
  • [*][*][*][*][*]
  • Developer (Donating)
[TOS #8] Coding for Public Distribution
Reply #23
Quote
randar, ffc, or whoever you are, if you wish to continue posting to this site (if you are not already banned under another alias) you will have to use a regular isp connection and not come to the site through an anonymous web proxy.


randar is ffc.  But ffc is banned from posting to this forum, so we won't see either ffc or randar here again unless ffc adopts both another persona and a new ISP.

ff123

  • fewtch
  • [*][*][*][*][*]
[TOS #8] Coding for Public Distribution
Reply #24
Quote
Originally posted by mpcfiend
Heh. Why is anybody still using MP3?

I mean, aside from the obvious portable uses. MPC is much much better for quality rips, and there is an MPC-MP3 transcoding tool available for portable users.

I would venture a guess -- for the same reasons people are still advocating weird Lame command lines, using the Blade encoder at 128kbps, calling 128k "cd quality," etc.  A combination of lack of knowledge, entrenched behaviours, and laziness about investigating alternatives.  The sort of combination that causes a sheep to remain with the flock at all costs.

BTW, transcoding sucks.  With 80 gig hard drives going cheap, I don't know why anyone bothers risking sound quality instead of re-ripping.

:coding:
Bring back dynamic range... www.loudnessrace.net