On Sun, May 4, 2008 at 10:23 AM, Rui Nuno Capela
<rncbc(a)rncbc.org> wrote:
Mark Knecht wrote:
<SNIP>
> ok.
we already have the "Root note", "Beats", "Meter" and
"Tempo"
fields in
sight,
which is a fair start imo.
Well, sndfile-info writes the word 'tempo' but I'm not completely
comfortable that we know where the tempo values are. Both of the
examples I provided say 120BPM. Unfortunately neither loop library is
actually recorded at that tempo!
that is *bad* news :( so the meta-data we get from libsndfile is bogus?
i guess we're back to square one (or is it zero?:).
No, I really don't thin kit's quite that bad. In fact I thought it
worked better than it did so maybe part of the problem is just lack of
attention over the last few years and some bug creeping in here or
there. Some of what's there is right, I think.
I really believe that just finding a couple of folks who have an
interest and digging in will probably figure it out pretty quickly.
<SNIP>
we'll have to resort that audio files must
be integral in length to ever
get some kind of loop auto-fitting ...
Well, independent of what the tempo is in a session, if I know an
8-bar pre-recorded loop is 120000 samples long (available from the
loop's meta-data) and I know my session requires 140,000 samples to
meet whatever the tempo is set to for 8 bars, then the resampling can
be done on the fly.
To be clear, Acid Pro does NOT resample on the fly with extremely high
quality. The quality is good but if I change a 68 BMP loop to 160 BPM
I am absolutely going to hear artifacts all over the place. My
experience is, however, that this gives 'character' to my work. It
sounds real, funky, fun. Most of the time what I hear when I listen to
a single resampled loop by itself is completely covered up when the
song is done. If it's not then and only then would I bother with
external, high quality resampling. 99.99% of the time it's just not an
issue for me.
but qtractor *does* time-stretching (what you call resample) on-the-fly!
you have two options here, being a trade-off between quality and cpu
intensiveness:
1) soundtouch based, wsola-like algorithm - very fast but artifact
prone--maybe with similar results as acid on-the-fly mode, you tell me
2) rubberband - good quality but resource intensive; might not be
appropriate for heavy loaded sessions and/or older cpu's
just to let you know. qtractor does *not* touch any of your audio sample
files, nor it creates any unless you're recording or exporting
one--almost everything is or can be done on-the-fly ;)
byee
--
rncbc aka Rui Nuno Capela
rncbc(a)rncbc.org