Re: [LAU] Timestretch without sample damage.

Hi Jonty,

I don't know on what algorithm the audacity and rubberband time-stretch
is based on (PSOLA, WSOLA, etc.... , time-domain or frequency domain or
whatever) but for what I know about that issue is that I never heard of
an algorithm that stretches a soundfile by putting silence in between
the extracted chunks. What about phase hits? ... the resulting signal
has to produce terrible click noises.

For example....

A simple PSOLA time-stretch algorithm would basically split the
audiofile *P*itch *S*yncronous into chunks of some ms, duplicate certain
chunks and reassemble the file by *O*ver*L*ap and *A*dd.

I stretched Rostropovich playing Bachs Cello Suite No 1 Prelude to -50%
and -99%. Didn't sound brilliant (lots of artifacts) but there were no
chunks of silence in between the samples --> i have to ask: what do you
mean by "into the samples" ? A sample is only one little value. Having
44.1kHz sampling rate means you have 44100 discrete samples per second.
There is nothing to put "into" a sample. Or do you mean the audio file
you used as an example?

Could you show us a spectrogram/waveform screenshot?

> Existing tempo changers that I've
> seen put some silence in the output file when I use them.

I never saw such a thing.

> Again, usual disclaimer: If I'm being stupid, let me know.

Why is someone stupid because of not knowing everything about anything?
Don't worry :)

Best regards,
Sebastian.
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
Received on Thu Jan 3 16:15:02 2008