9 Mar
2007
9 Mar
'07
6:13 p.m.
On Fri Mar 09, 2007 at 11:09:27PM +0100, Robin Gareus wrote:
terrence(a)terrencevak.net wrote:
theres plenty of software that can take a FFT snapshot, wiht a large window size (eg a
passage of spoken text). then filter another wave using that as a mask...
you could also play with a spectral gate, if the voice is obviously louder than everything
else. once again you'll get artifacts and sectinsn that didnt hit the threshold..
it definitely helps to work in an editor that lets you view your waveforms as a
spectrogram as well.
theres lots of choices on windows for both of these things (fft masking, and spectrogram
editing). not sure about linux..
Hmmmm...well it's actually an old tape of
all spoken stuff;
conversations I had while visiting family in Europe. I want to hang on to
the other people's voices etc., but mine has changed a bit since then and
the old sound bugs me.
i don't think there is a way to do this from the command line!
why cli? it rather sound like a job you should start up a (graphical)
DAW for. or was the cli requirement only for the recording not this
particular edit/filtering task?
anyway the only way to get it done quickly, just mute your voice during
recording.. (hope you were not conversing at the same time as others in
the background) - Maybe ecasound allows to assign on/off location
markers during recording. I never checked.
If you speak at the same time with others it's gets tricky. If the
voices have a different pitch (fi. your "old sound" being an allusion to
baby-crying in the background) you can reduce the presence of a
particular voice with EQing + compression. I don't think there is
software out there to automate the task. - audacity might come in handy.
if you want to get into cli denoising based on a audio-sample:
try http://www.saunalahti.fi/%7Emjkoskin/anoi-0.3.tar.gz
robin