Following the recent debate 'tweaks vs presets' I collected some of the tweaked lines that people posted on this forum. And I tried to put up a linear q-scale, a la oggenc.

I will modify it if someone has more tweaked lines to add, or if some lines are in a wrong place on the scale. Sort of a draft proposal from users to developers.
Then it is up to Frank whether to use it somehow or ignore.

And yeah, as I logged in to the forum to post it, there it was - a message from Frank about new 1.05. It rocks!!!

Well, I guess I'm late , but I post this anyway. Still it could be useful.

==================================================================


q > 10 --braindead --nmt 15

q = 10.0 ? --braindead

6.0 < q < 10.0
--insane --minSMR 0 --nmt 16 --tmn 32
--insane --minSMR 0 --nmt 12 --tmn 32
--insane --minSMR 0 --nmt 12 --tmn 28
--insane --minSMR 0 --nmt 12 --tmn 26
--insane --minSMR 0
--xtreme --nmt 12 --tmn 28

q = 6.0 --extreme

5.0 < q < 6.0
--xtreme --ltq ank
--standard --ltq fil
--standard --nmt 8

q = 5.0 --standard

4.0 < q < 5.0
--xtreme --nmt 7 --tmn 19 --ltq iso --ltq_gain -15 --ltq_max 70 --bw 18000
--xtreme --nmt 5 --tmn 14 --ms 1 --ltq iso --ltq_gain -10 --ltq_max 77 --bw 17500
--standard --bw 15000

q = 4.0 --radio

q = 2.0 ? --thumb

q = 0.0 ? --telephone


This q-scale is strictly quality oriented, so no --insane full frequencies.
The insane switch should either be gone for good, or modified and renamed to something like
preset 'transcodable'.

================================================================


--transcodable

Must satisfy the condition that transcoded mp3/ogg sound the same as direct encoded ones at some reference mp3/ogg bit-rate (128 kbps?). That is, to sound the same to an average listener, no golden ears.
A whole lot of ABX'ing to work it out.

This could be defined as a reference "transcodability" level = 0.

Then one could perhaps introduce a "transcodability" scale varying from -10 (barely listenable transcodes) to +10 (ultimate quality -- transparent for golden ears at ANY possible mp3/ogg bitrate)

Should it be based on insane? or braindead?

Some 'insane' lines people using:

--insane --nmt 16 --tmn 32
--braindead --minSMR 3

Regards,

Konstantin

(
And what the heck is this line good for?
--insane --ans 0 --ltq_gain -12 --nmt 15 --tmn 34 --shortthr 2
)

P.S. Updated the q=4,5,6 according to new v1.05e settings.
And I think I'll stop here. After 1.05 is released there's not much sense in making a parallel q-scale.

1.05 any more news ?

Judging from what Frank Klemm posted, "--quality 5.0" will be equivalent to "--standard". Since MPC doesn't perform best at low bitrates, we don't need that many steps for the lower end (in your post that would be 0.0 to 8.0, too much).

Yes, I thought about it. What I had in mind is to make some headroom for future improvement of low bitrates. Is it planned? or possible at all?
Because if it's not possible to improve, and nobody use it then....


And another thought was that visually q value = 5 out of 10 suggests a rather mediocre quality ad hoc. But --standard is already a very high quality. Sort of contradiction to what I (personally) expect from such a q value.
Does it make any sense?

Comparing to the ogg scale, where should mpc-standard go? something about ogg -q7 ? (comparing quality, not the bitrate)

That is, I put standard on q=8 for psychological reasons. If it's only me alone feeling that way, I surely change it.

How about --standard q = 7 as a compromise?

First, you have to realize this: The tables are most likely almost completed. So this remains a theoretical discussion.

Nonetheless, let me throw me 2 (Euro-) cents in here too. I think that the "--quality 5.0" = "--standard" is a good decision. After all, people expect the default setting to make most of their music sound good. Think LAME --aps.

It's justified to put that as --quality 5.0, and even more so because Ogg can't deliver in the same way with it's -q 5... maybe with -q 7, as you said. Even more proof that MPC is the best medium- to high-bitrate solution right now, and also an encouragement for the Ogg Vorbis folks to improve on the higher bitrates, i hope.

So, after all, MPC --standard was and is the starting point, and people can go up and down from there, to fit their needs.

Yes, I already wrote I was late -- after Klemm announced 1.05.

It is not to do Frank's job or to make some demanding request. It's just for us users to think of what we users would like to see with mppenc, perhaps in a remote future. Theoretically, I agree.



What can stop us from making --standard = q=7 to be default?
That is, q=5 doesn't have to be default value in the scale.



OK, I'll change it to q=5.

What I really wouldn't like to happen is that new users (and not so new too ) will encode everything at the setting q=10, just because on the 0-10 scale it looks "better quality" than q=5.

What we have now: despite standard being the default preset, majority of people skrew on it and use extreme, insane and whatever makes "better quality". They will skrew on default q=5 value too.

You're going to have that anyway. But I'd guess that the vast majority of people are at least going to try the default setting to see how it sounds before trying anything else.

Unless they've got friends in this forum that are going to say "dude, -q 5 is WAY too low. Encode at -q {7,9,11}" In that respect, not many people here really accurately represent "average" users.

Overall, I agree with CiTay, --standard has been the target for mpc, and -q 5 should represent the "average" quality that you're going to get out of the encoder.

Besides, if you want to play psychology tricks, I'd map --standard to -q 3. People are going to try mpc on it's reputation as a superior codec., and for them to hear something at what looks like the lower end of the spectrum sound so good might really impress them. But that's just me.

"It makes much bigger files than Ogg and Ogg sounds just as good! It sucks!"

(Majority of people think -q3 is transparent)

--
GCP

or put a nasty disclaymer that q5 is transparent. (If someone thinks it's not then hes a moron / vau, congrats you found a sample where mpc fails(very slightly) please send it to pfk at uni-jena.de and you get a 1000$ award) :insane: ???

Edit: added smilies for serious people.

Don't you think it is something rather weird and absurd what you're suggesting?


I'm afraid you totally missed the point I was trying to make.

The q-scale is supposed to reflect the perceived sound quality level. It should assist the users to choose the right setting rather than trick them into anything.

When a very high quality preset such as standard is put right in the middle of the scale, it gives me a false impression that such preset is mediocre quality, with a lot of headroom for quality improvement (from 5.0 to 10.0). But in reality there will be no improvement in perceived quality. Standard is already transparent for me.
This was the 'psychological' reasoning I was trying to show.

Excuse my poor English.

[Edit typos]

I understand the point that you're making, but I think that it only makes sense when trying to compare two fundamentally different quality scales to each other. It makes sense for mpc's quality scale to have standard at -q 5, as we've both agreed upon earlier. People are estimating this to be closer to -q 7 on the vorbis quality scale. And it would probably place it closer to -q 8.5 on a similarly designed scale for lame.

This all makes sense, because mpc was designed to have higher quality than these other encoders. It's "standard" -q 5 is much higher quality than vorbis or mp3 at the same level. (As I mentioned in another thread, if you plugged -q 5 into mpc, vorbis, and lame, you'd get three very different sounding files. ).

The vast majority of users who download and use mpc understand that it's designed and held to a higher standard. Anyone else who wanders across the site at random will find out really fast.

I hope I'm still making sense.

Dibrom, for instance, already found a few. Will he get $1000 for each one, or in total?

My apologies. I forgot the damn smilies:mad:

I was just kidding, of course...


:too_hot:

Daniel, relax. You're pretty much mistaken thinking I didn't get it was a joke of yours.
Just wish you could have said something on topic of the thread in addition...


Btw, I usually don't type plain text emoticons myself either. But these little bastards :eek: etc are so nifty - just couldn't help using them sometimes!