Posted by fred-bloggs on 01/19/06 11:06

Wes Groleau <groleau+news@freeshell.org> wrote in
news:pnDzf.17979$sq.6732@trnddc01:

> fred-bloggs wrote:
>
>> --comp should work, but I suspect that Lame calculates a bitrate from
>> the required compression ratio:
>> b = 1411/comp (1411 is CD bitrate)
>> and then chooses the *next* standard bitrate.
>> You would have more control if you specify a standard bitrate
>> directly.
>
> Makes sense. But the documentation implied that you when get the
> compression you specified or better. Not so.

The USAGE document with my version 3.97 doesn't even mention --comp.

>> You may be thinking of variable bitrate encoding (VBR) where Lame
>> encoder reduces the bitrate when encoding silence or simple tones.
>
> Yes. The documentation also implied that VBR was used unless
> explicitly set otherwise. Or maybe I interpreted it wrongly.
>
> -b 32 --resample 11.025 --highpass 5
>
> gave me the same size ratio as
>
> --comp 21 and --comp 25

Earlier you said that --comp 44 gave you 900 MB.

>
> which was approximately 22.07 and would make the
> whole book approximately 900 MB. (I'm aiming for
> a single CD.) I left out the -a because I had already
> used 'sox' to average the channels into mono.
>
> (Also, for what it's worth, using -a and --comp
> at the same time crashes this version of 'lame')

If you need to convert to mono again try specifying the mode as mono,
viz:
-m m

>
> I am going to try -b 24 and -b 16.

24 should give you about 675 MB, 16 about 450 MB.

> I am also going to try other sampling rates.

I suggest you use sample rates related to 44.1 khz (22.05 or 11.025), it
makes the resampling a lot simpler.

> Are you sure about highpass ? I would expect lowpass.
> (If you can call 5Khz "low")

*lowpass*, my bad, sorry.



--
fred

[Back to original message]