I haven't heard the effect, and phase randomization to achieve an "infinite stretch" or "infinite reverb" is definitely a useful thing, but I don't think this is something we should advertise as time stretching. I would expect at least a correctly implemented phase vocoder under the label "time stretch." As long as we don't call this "time stretch," I have no objections to it. Getting a good time stretch effect is another issue.

-Roger

 
On 5/26/12 3:48 PM, Paul N wrote:
Hi Michael.
I updated the patch with 3 spaces instead of tabs and attached into the email.

Regarding people wanting this, the effect became quite popular in last few years. Here are some articles about it:

http://www.washingtonpost.com/lifestyle/style/time-stretching-warps-songs-into-sprawling-blurs/2011/12/27/gIQASnwARP_story.html

 http://www.g4tv.com/videos/55210/pauls-extreme-sound-stretch-play-with-sound-review/

 http://www.npr.org/blogs/therecord/2010/08/18/129283985/the-art-of-a-time-stretch


Regarding the algorithm, you can see better explained here (with minor modifications than the code submitted to Audacity):
http://zynaddsubfx.sourceforge.net/doc/paulstretchsteps.png

Indeed, it is a FFT->do_something->IFFT approach, but the phase randomization is an important step. It sounds best on choirs ( like this one  http://ericwhitacre.com/blog/the-virtual-choir/virtual-choir-2-0-slowed-to-61-minutes ) , ensemble music (like this one http://soundcloud.com/birdfeeder/jurassic-park-theme-1000-slower  - this has over 1 million views) .

It is not just useful for time stretching but also to smear the sound over a period of time (if you set stretch=1.0 and window to a large value, like 5 seconds). This is an useful tool to create ambient music.

Paul