Since everyone's hearing is different, your best bet is to try a number of bitrates and try ABXing them with foobar2000, or any other handy ABXing program.

ABXing is comparing two files (A and B) to the reference X. If you can repeatedly successfully identify whether A or B is X, the files have a great discrepancy in quality. Increase the bitrate until you can't reliably ABX it with the one before it.

By the way, "-V2 --vbr-new" is a good recommendation, but "-V2 --vbr-new -q0" is of higher quality. Both the speed and the quality increase is small. The default is "-q5".

EDIT: Alternately, you could use "--preset fast standard -q0". And yes, "-q0" is a pet peeve of mine

UED77

UED77,

I'd like to see some ABX tests where q0 makes a difference for you.

Where do you think does the q0 switch together with -V2 --vbr-new improves most in your listening experience?
At least when using the actual hydrogenaudio recommendation with -V2 --vbr-new setting q0 is totally ignored.
So it makes me wonder it even makes a small increase in quality

First of all, like you were told, every person has different hearing capabilities. For me LAME reaches transparency with the VBR setting 'lame -h -V4 --vbr-new' (cca. 160 kbps avg.) or with the CBR setting 'lame -h --alt-preset cbr 160'.

When using VBR new the -q settings work differently, and in fact -q0 will do exactly the same thing as the default quality setting (3, I believe).

I did a quick test using --verbose, and the only difference I could see in the report (using WinMerge) is below:

-V2 --vbr-new


-V2 --vbr-new -q0


Does this mean anything?

As you say, -q3 is used in the first run. Files were all bit-identical.

Edit: Similarly -h -V4 --vbr-new (post #4) produces the exact same results as -V4 --vbr-new. -q2 is used over -q3, but this appears to have no effect. The settings reported by --verbose are totally uneffected by this switch.

-q3 or -q0 are producing the SAME file with vbr-new.
Under vbr-old or cbr/abr, the difference is the noise shaping amplification mode (ie how do you gradually increase scalefactors to match the masking values). As vbr-new is using a more direct method, this is irrelevant.

A good example of placebo effect.

After I wrote the post, I did become afraid that I might be slammed by a TOS #8, but the thread split really surprised me.

I was under the assumption that "-q5" was the default. It is stated so in the lame -? documentation.

I performed three tests, "-q0", default, and "-q5". It turns out, the default is, in fact, "-q3". I know I should pay more attention to the command line.

q0 and q3 do produce the same file. q5 produces slightly larger files, and tends to use more higher-bitrate frames than q3.

I was wrong. Now I need a hug

UED77

Thank you for you post UED77.

You have just had first hand experience of Hydrogen Audio's no-nonsense approach to misinformation.

TOS 8 is "the very core of Hydrogenaudio", therefore any claims regarding the improvement or detriment of an audio source must be based in fact only, and not "pet peeves".

The thread was split in accordance with TOS 5, as the posts regarding your post were completely unrelated to the thread's topic. Splitting threads in this way ensures that a thread stays on track, and can be used as a document regarding a specific question. If a thread is allowed to miander it makes it very difficult for future readers to get the information that they need.

Both ethos ensure that the information here is as accurate and clear as possible.

Hopefully the experience will only go to show you that the members here are serious about their audio, and not that new members are not treated well.

Welcome to Hydrogen Audio.