14 Dec
2007
14 Dec
'07
5:13 a.m.
[Chris Cannam]
On Monday 10 December 2007 23:36, Tim Goetze wrote:
Indeed, the chorus/aliasing effect is as good as gone with
--no-peaklock, but the vocals begin to suffer from periodic amplitude
modulation (tremolo). The effect even becomes a bit stronger with
--crispness 5.
Unfortunately it also has to be noted that
after the onset, the
voice body in rubberband's output is not sounding quite as good as
with 'stretch' -- there's a faint chorus/aliasing effect, not very
strong but irritating.
You may find you prefer it with the --no-peaklock option, or even
with --crispness 5. I perhaps foolishly didn't include a standard
crispness option that only
does the equivalent of --no-peaklock, because that wasn't preferred to
any of the alternatives for the examples in my informal blind listening
tests. (Instead the default settings tone down the amount of peak
locking gradually as the stretch ratio increases.)
At what stretching/compressing ratios have you run your tests, and
with what kind of source material? While the code is certainly doing
quite fine, I have a feeling that my 2x test runs could be a bit out
of line with the intended kind of use.
Tim