Am Sonntag, 27. Februar 2011 17:08 schrieb Rogério Brito:
> What bitrates do you want to achieve, in particular? Something like:
> | n | avg. kbps |
> | 0 | 256 |
> | 1 | 224 |
> | 2 | 192 |
> | 3 | 160 |
> | 4 | ?? |
> | 5 | 128 |
> | 6 | 112 |
> | 7 | 96 |
> | 8 | 64 |
> | 9 | 48 |
Once it was like 260, 220, 190, 170, 150, 130, 112, 96, 80, 64, 32
now more like ..., 115, 100, 85, 70, 35
> With this moving of mine, I will quite probably be ripping a lot of my
> collection and be encoding stuff. I can report what I have as a result.
> My collection mostly has heavy metal stuff, which should, in some sense,
> suffer from bloat of the sfb21 problem. :-/ (Especially Black Metal stuff
> with lots of high frequencies).
I guess your collection will result in higher average bitrates then.
> > Another topic right now is, CBR/ABR encoding was reported of being
> > not so good, as of lately. As a quick fix they are using the VBR
> > psy model. Adapting the internal parameters to make the best use
> > of it is somethings I'll have to do next.
> I see that you changed the vbr methods back and forth from/to vbr-mtrh,
> right? BTW, could we have a brief paragraph on how the methods differ?
I used the then unused switch --vbr-new (internal name vbr_mt) for tunings
of the VBR psy model. Now that I'm done, --vbr-new and --vbr-mtrh are
the same again, utilizing latest results.
> If I understand it correctly, MT and RH are suposed to be the initials of
> Mark Taylor and Robert Hegemann, right? :-)
> > One thing that would be nice to have in 3.99 is, fixing the problem that
> > the shared libraries do export a lot of internal stuff. Like the Windows
> > DLL, the dynamic link library should export symbols from lame.h only.
> This means that, since we are changing the API, we should have to bump the
> library SONAME, right?
Theoretically, nobody should have used library internal symbols and bumping
SONAME shouldn't be necessary. But, we should bump after hiding the internal
symbols. And when we bump SONAME, we eventually can get rid off some
deprecated API functions too.
> Regards, Rogério Brito.