On 06/26/2010 08:32 PM, Chris Cannam wrote: > On Sat, Jun 26, 2010 at 4:03 PM, Olivier Guilyardi <ml@xxxxxxxx> wrote: >> Rubberband is interesting, but when I had to develop a professional tool to time >> stretch voice recordings, the quality was far from acceptable. > > No, Rubber Band is better for some sorts of audio than others and > speech certainly isn't its strong point. Yep, that's right. I did a quite lot of tests, and realized every stretching lib is better for some sort of sound than others. Also, some are optimized for speed, not quality. I've actually been happy with Rubberband for my own music stuff, but when I had to develop a critical tool for mass stretching of voice recordings, Dirac just did an incredible job. As you may know, it seems to be the result of decades of developement by the founder of Prosoniq, which also produced MPEX. > Tim Goetze also had an example recently (here or on LAD) which it > didn't handle well -- I've been looking into that one recently and > hope to have some improvements there for the next release at least. Also, when I mention Dirac support in Ardour, I don't mean that it should replace Rubberband. In such apps as Nuendo and the like, the user has a choice about which algorithm to use. Plus, since Dirac isn't FLOSS, it wouldn't be installed by default in many cases I suppose. AFAIK, Rubberband and SoundTouch are still the only free stretchers in the market place :) But, from my tests, I believe that you could improve things a lot in regard to voice processing. I can recommend that you take almost any piece of voice, stretch in the 80%-120% range, and compare the Dirac result with Rubberband's. -- Olivier _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/listinfo/linux-audio-user
