Hello to all.
Using the alt preset(s) in the Lame 3.92, i must say that
"alt preset 128 cbr" make a good job avoiding common problems
at this bitrate.
Can someone explain the internal switches or commands used by
alt preset 128 and alt preset insane?
And, if possible, the meaning?

I'd recommend --alt preset standard; generally if standard fails (VERY FEW samples) extreme or insane will show little to no improvement. or --alt preset 128 (abr) if you need bitrates very close to 128.

Only Dibrom (the presets developer) can answer what the internal switches really do.

HINT: use the SEARCH function once in a while

--alt-preset 128 is:

--alt-preset cbr 128 is otherwise the same except --abr 128 is replaced by -b 128

About the meaning.. well most switches are pretty basic, --athtype 2 chooses more precise absolute threshold of hearing -curve, --ns-bass -6 will give more bits (higher resolution) for low frequencies.

--alt-preset insane : It uses code level tweaks, so you can't find switches which would be identical to it. You can see basic properties with --verbose switch.

Verbose? Sure!
Thanks a lot for the support.

1) I thought if you used, say --alt-preset cbr 128, it was the same tunings as the --alt presets only "locked in" to 128

2) I didn't know scale was involved. Is it also involved in aps, api, ape?

What do you mean "same tuning"? Code level tuning is used with --alt-preset standard,extreme and insane. Other --alt-presets are implemented with different normal switches.
Scale is not used in standard,extreme or insane -presets.

Well, I surely thought there was some code tweaking also in:
--alt-preset <bitrate>
I thought some switches were affected depending on the bitrate...
(For example, I noticed that if you go too low, you'll get audio resampled to 32Khz...)

Dibrom had plans to introduce some of the tweaks of aps into the ABR presets...

(I hate to ask this, as I certainly do not want to even mildly put any kind of pressure on him, but...) Is there any progress in that? Or all the other projects are keeping him too busy?

It was also nice to hear him say he tried some of the Naoki experimental code and got bitrate to below r3mix levels keeping nice quality... I wonder when that will see public light...

I commonly use --alt-preset between 135 and 145 for my portable needs...

Lame resamples, if you use a bitrate (abr/cbr) below 112 kbps.

dev0

Sure, there are different switches used for different bitrates, but those are not exactly code-level tweaks. Those are just normal preset alias settings implemented with different switches and bitrates, but which do not include actual code level tweaks like standard,extreme and insane.

Things that change depending on the bitrate are lowpass-values ,ns-bass values, nsmsfix values etc..

But the actual code level tweaks deal with:
- Tweaked block switching threshold
- Adaptive noise measurement (uses X3 when needed)
- Tweaked noise shaping functions
- Short block tweaks (bitrate, nsmsfix values)
etc.etc. which you cannot set with normal switches.

Once again, the hordes of experts lurking in HA rise to the rescue of a poor n00b...

... and, alas, he is retrieved from the mist of darkness of his own ignorance!

Thank you both for your quick reply

I will keep my hopes high for an even higher quality 140-150 bitrate preset...

(the current ABR presets are perfectly good enough for my ears, and surely my Archos loves them -loved them until it died, it's waiting for soldering surgery to be performed on it and be brought back to life... - but it's good to know you can squeeze more quality of the same bitrate range...)

I actually have wet dreams of 2-pass audio encoding...

I wanted to know what lowpass is used for standard/extreme/insane.

standard lowpasses at 19 kHz, extreme at 19.5 kHz, and insane at 20.5 kHz.

Thanks a lot for the support!

I searched the FAQ and WIKI some more and come up with the internal preset extreme settings for LAME 3.95.1
and for LAME 3.96.1
I would like to know what the settings in LAME 3.90.3 are, since its encoding window shows VBR(q=2) ??? Isn't is supposed to show q=0? (I know, it's just a detail )

Hello, welcome to HA.

I see that you searched before asking, that is very good.

The truth is that the presets use code-level tweaks that can't be replicated using switches That reference you show is only that, reference.

As for q 2 or 0, that has been discussed many, many times (also, the hmm, "inconvenience" of preset extreme, have you tried standarr? you would save a lot of space and if you can hear a difference you will have the undying respect of a lot of people around here). Search for an answer. Short one: q 2 is more than you can possibly need.

Welcome again.

Thanks for the warm welcome

I am going to backup my CD collection and I would like to do so in very high quality - however, CBR 320 seems overkill to me. So I go for the tested LAME 3.90.3 and --alt-preset extreme.

I tried later versions with VBR and q=0. My question is, how --alt-preset extreme in 3.90.3 compares to 3.96.1 and the setting -v -V 0 -q 0 ?

@Kitanai: First, there's no need to specify -v when you're already specifying -V0 (which is equivalent to --preset extreme, as it seems you've already discovered).
Second, I (and most others in these forums) would recommend against specifying a different -q value when already specifying a preset VBR setting. The people who set these up know what they're doing, and specifying -q 0 isn't going to improve the encode. More generally, even if you're designing your own commandline switch, -q0 might not be the way to go... the Lame helpfile says the following:

Finally, the -V settings weren't implemented until 3.95, so -V0 won't work in 3.90.3, or at least it won't get you the equivalent of --preset extreme.

Most people on here would recommend that you use 3.97a11 -V0 --vbr-new (a.k.a. --preset fast standard) which will probably be slightly better in quality - and a whole lot faster - than Lame 3.90.3 --preset extreme. Lame 3.90.3 was specifically optimized for --preset standard, and is still recommended specifically for that switch, because it's been tested so much. (Even so, 3.96.1 is used by more people, and 3.97a11 used by even more people and has tested extremely well when using the --vbr-new switch.)

Thanks for your thoughts on this issue - it basically supports my decision. I guess I just stick to 3.90.3 since I already encoded quite a few files with its extreme setting. I little extended encoding time won't hurt me 2.8 GHz machine

Thanks again!