30 Mar
2009
30 Mar
'09
8:12 p.m.
I've been trying to do some sound morphs between two different hits of a snare drum,
one rimshot and the other a plain hit.
I tried using SMSTools, and even some windows programs under wine, but the results always
seem to sound more like hitting a bucket full of water than a combination of the two
drums.
I tried csdrummer, but only get a horrible scratchy noise from it.
I looked at Loris...the python-based example that comes with it seems horrifically
complicated.
Maybe someone with a knowledge of Csound/Loris could be kind enough to help with a simple
script? Or a step-by-step how to achieve this with Cecilia?
Are there any other programs I could try?
30 Mar
30 Mar
8:38 p.m.
hi shane,
i've done a python script to do some morphing using wavelet decomposition of
two sounds.
it was done for experimental pourposes and i liked its results very much.
you have to tweek the script a little bit to use your samples.
actually substititute s1 and s2 by your samples with
sf= A..sndfile("sound1.xxx")
s1=sf.read_frames(sf.get_nframes())
and match their sizes with zeros or with wavelet coefficients already on
decomposition step.
note that it outputs different morphs of the two sounds. and that you can
remove the import pylab statement from the header as the script is not using
it. its here:
http://trac.assembla.com/audioexperiments/browser/pya/morph.py
hip-hip-hurray,
renato
2009/3/30 shane richards <shanerich(a)email.com>
I've been trying to do some sound morphs between
two different hits of a
snare drum, one rimshot and the other a plain hit.
I tried using SMSTools, and even some windows programs under wine, but the
results always seem to sound more like hitting a bucket full of water than a
combination of the two drums.
I tried csdrummer, but only get a horrible scratchy noise from it.
I looked at Loris...the python-based example that comes with it seems
horrifically complicated.
Maybe someone with a knowledge of Csound/Loris could be kind enough to help
with a simple script? Or a step-by-step how to achieve this with Cecilia?
Are there any other programs I could try?
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
8:44 p.m.
hi shane,
i've done a python script to do some morphing using wavelet decomposition of
two sounds.
it was done for experimental pourposes and i liked its results very much.
you have to tweek the script a little bit to use your samples.
actually substititute s1 and s2 by your samples with
sf= A..sndfile("sound1.xxx")
s1=sf.read_frames(sf.get_nframes())
and match their sizes with zeros or with wavelet coefficients already on
decomposition step.
note that it outputs different morphs of the two sounds. and that you can
remove the import pylab statement from the header as the script is not using
it. its here:
http://trac.assembla.com/audioexperiments/browser/pya/morph.py
hip-hip-hurray,
renato
2009/3/30 shane richards <shanerich(a)email.com>
I've been trying to do some sound morphs between
two different hits of a
snare drum, one rimshot and the other a plain hit.
I tried using SMSTools, and even some windows programs under wine, but the
results always seem to sound more like hitting a bucket full of water than a
combination of the two drums.
I tried csdrummer, but only get a horrible scratchy noise from it.
I looked at Loris...the python-based example that comes with it seems
horrifically complicated.
Maybe someone with a knowledge of Csound/Loris could be kind enough to help
with a simple script? Or a step-by-step how to achieve this with Cecilia?
Are there any other programs I could try?
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
31 Mar
31 Mar
4:52 p.m.
2009/3/30, shane richards <shanerich(a)email.com>om>:
I've been trying to do some sound morphs between
two different hits of a snare drum, one rimshot and the other a plain hit.
I tried using SMSTools, and even some windows programs under wine, but the results
always seem to sound more like hitting a bucket full of water than a combination of the
two drums.
Drum sounds are very "difficult" for these tools because the FFT-based
analysis assumes a quasi-stationary signal (imagine a sine wave that
never changes, that's the easiest) whereas drum hits really strongly
violate this. Also phase issues and other stuff. To be honest you'll
probably do best using some kind of clever perceptually-tweaked
crossfade-based technique... or maybe find something using LPC-based
analysis rather than FFT, that might do well (I haven't looked) - LPC
should have a better chance at handling drum sounds gracefully.
Dan
--
http://www.mcld.co.uk
6:28 p.m.
Shane: that script was done using pywt version 0.1.6 (alpha) i think that
its current. what version did you find?
Dan: wavelets are used to represent/compress or synth signals with
transients. kind of opposed to FFT to stationary signals... i dindt use LPC
but i plan on studying it, do you know any good material?
2009/3/31 Dan S <danstowell+lxau(a)gmail.com <danstowell%2Blxau(a)gmail.com>>
2009/3/30, shane richards
<shanerich(a)email.com>om>:
I've been trying to do some sound morphs
between two different hits of a
snare drum, one rimshot and the other a plain hit.
I tried using SMSTools, and even some windows
programs under wine, but
the results always seem to sound more like hitting a
bucket full of water
than a combination of the two drums.
Drum sounds are very "difficult" for these tools because the FFT-based
analysis assumes a quasi-stationary signal (imagine a sine wave that
never changes, that's the easiest) whereas drum hits really strongly
violate this. Also phase issues and other stuff. To be honest you'll
probably do best using some kind of clever perceptually-tweaked
crossfade-based technique... or maybe find something using LPC-based
analysis rather than FFT, that might do well (I haven't looked) - LPC
should have a better chance at handling drum sounds gracefully.
Dan
--
http://www.mcld.co.uk
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
6:32 p.m.
sorry:
In [1]: import pywt as W
In [2]: W.__version__
Out[2]: '0.1.7'
(svn version afaiRemember)
2009/3/31 Renato Fabbri <renato.fabbri(a)gmail.com>
Shane: that script was done using pywt version 0.1.6
(alpha) i think that
its current. what version did you find?
Dan: wavelets are used to represent/compress or synth signals with
transients. kind of opposed to FFT to stationary signals... i dindt use LPC
but i plan on studying it, do you know any good material?
2009/3/31 Dan S <danstowell+lxau(a)gmail.com <danstowell%2Blxau(a)gmail.com>>
2009/3/30, shane richards <shanerich(a)email.com>om>:
I've
been trying to do some sound morphs between two different hits of a
snare drum, one
rimshot and the other a plain hit.
I tried using SMSTools, and even some windows
programs under wine, but
the results always seem to sound more like hitting a
bucket full of water
than a combination of the two drums.
Drum sounds are very "difficult" for these tools because the FFT-based
analysis assumes a quasi-stationary signal (imagine a sine wave that
never changes, that's the easiest) whereas drum hits really strongly
violate this. Also phase issues and other stuff. To be honest you'll
probably do best using some kind of clever perceptually-tweaked
crossfade-based technique... or maybe find something using LPC-based
analysis rather than FFT, that might do well (I haven't looked) - LPC
should have a better chance at handling drum sounds gracefully.
Dan
--
http://www.mcld.co.uk
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
3 Apr
3 Apr
10:19 p.m.
On Tue, Mar 31, 2009 at 2:52 PM, Dan S <danstowell+lxau(a)gmail.com> wrote:
or maybe find something using LPC-based
analysis rather than FFT, that might do well (I haven't looked) - LPC
should have a better chance at handling drum sounds gracefully.
Yeah, or some granular approach; interleaved samples, very short,
while first sound's density decreases, the next sound's density
increases. I think that if I were doing morphing of busy spectra I
would go with that first (but I have never tried this so don't take my
word for it). Needless to say, I would not advise to do the splicing
by hand ;)
--
./MiS
514-344-0726
http://www.creazone.ca
5509
days inactive
5513
days old
linux-audio-user@lists.linuxaudio.org
6 comments
4 participants
participants (4)
-
Dan S -
Michal Seta -
Renato Fabbri -
shane richards