Announce: Fast real-time-stretching and resampling algorithms (Procrustes/Sylea)

Started by DSP team May 21, 2006
DSP team wrote:
>>Did you try different content? I tried a couple of orchestral things, >>and it sounded pretty good. Then I tried a piano sonata. Speedup there >>sounds OK, but even small amounts of slow down sounds awful. It sounds >>very watery, like the algorithm really can't get a grip on the pitch. >>That said, the orchestral bits I tried were pretty neat. > > > This is very strange, because I hear nothing of the kind with the > Beethoven's Moonlight sonata (available for download for example at >
http://multimedia.utsa.edu:16080/technology/students/021/JVanAntwerp/Moonlight_Sonata.mp3)
> - it sounds quite normal to me, including when slowed down. > > What is the sonata you tested? Is it available for download? > > Audio DSP team
I can't remember which exact sonata it was that sounded awful, but it was Mitsuko Uchida playing one of Mozart's. I have the whole set of Mozart Sonatas on my notebook, losslessly ripped from my set of CDs. I picked a few odd files at random from those, and from a set of Mozart concertos, to try. The really bad solo piano pieces were medium speed, and not desperately complex. Regards, Steve
>I can't remember which exact sonata it was that sounded awful, but it >was Mitsuko Uchida playing one of Mozart's. I have the whole set of >Mozart Sonatas on my notebook, losslessly ripped from my set of CDs. I >picked a few odd files at random from those, and from a set of Mozart >concertos, to try. The really bad solo piano pieces were medium speed, >and not desperately complex.
All right, I went and downloaded all available 30-second MP3 samples at a New Zealand site offering Mitsuko Uchida's performance of Mozart's piano sonatas (http://www.opuscds.com/cd/29490) and played them all at 75% tempo. Again, I didn't notice anything bad. They're sampled at a low rate indeed, but for the piano sound that may be acceptable, as it doesn't contain too high frequences. Could you please try to find the bad sounding file, pack it with WAVPACK or LAME, and make it available somewhere for download (or as a last resort, e-mail it to us)? You could also extract a section which sounds bad to you. By the way, did you test it with the Winamp plugin or the demo player? If it was the player, which version[s] did you test? Regards, Audio DSP team
Steve Underwood wrote:
> Bob Cain wrote: >> DSP team wrote: >> >>> Instead of shifting the pitch and resampling, you can do it at once! >>> E.g., >>> there is a real-time-stretching algorithm called "Procrustes" >>> available at >>> our web site - http://www.dspteam.com/ - with demo players showing how >>> fast yet good sounding it can be done if written 100% in Assembler and >>> optimised to extremes by utilising the x86 SIMD instruction sets 3DNow! >>> and SSE while taking care of the transients. It's available for >>> licensing. >> >> Gotta compliment you on the sound quality of your time/tempo >> stretching demo players. >> >> I would also be interested to hear players that do real time >> frequency/pitch scaling. >> >> A more complete demo would be a player and plugin that does both >> scalings simultaneously controlled by 2 sliders having time/tempo on >> one and frequency/pitch on the other. >> >> >> Bob > > Did you try different content? I tried a couple of orchestral things, > and it sounded pretty good. Then I tried a piano sonata. Speedup there > sounds OK, but even small amounts of slow down sounds awful. It sounds > very watery, like the algorithm really can't get a grip on the pitch. > That said, the orchestral bits I tried were pretty neat.
I tried it on two different types, one mostly percussion with quite a few different types of instrument, and the other Diana Krall doing vocal over piano. I was impressed at the larger scalings and found smaller ones quite transparent. What sonata did you find problems with? Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein
DSP team wrote:
> > There are spectra for both 440 Hz and 16 kHz sine waves there. While 16 > kHz isn't the worst case, it's close.
You missed my point. My point is that the worst case is when you are performaing downsampling and the source signal has significant amounts of audio energy about the nyquist frequency of the destination sample rate. Erik -- +-----------------------------------------------------------+ Erik de Castro Lopo +-----------------------------------------------------------+ "Safety versus Expressiveness is a false dichotomy -- you can have both. Compare ObjectiveCaml with CeePlusPlus: OCaml obtains expressiveness without compromising safety, while C++ obtains it by throwing away safety. The latter is just bad design." -- David Hopwood
DSP team wrote:
> >I can't remember which exact sonata it was that sounded awful, but it > >was Mitsuko Uchida playing one of Mozart's. I have the whole set of > >Mozart Sonatas on my notebook, losslessly ripped from my set of CDs. I > >picked a few odd files at random from those, and from a set of Mozart > >concertos, to try. The really bad solo piano pieces were medium speed, > >and not desperately complex. > > All right, I went and downloaded all available 30-second MP3 samples at a > New Zealand site offering Mitsuko Uchida's performance of Mozart's piano > sonatas (http://www.opuscds.com/cd/29490) and played them all at 75% > tempo. Again, I didn't notice anything bad.
There is quite a difference between the Fast and PRO modes. For piano (Erik Satie, Gnossienne), the Fast mode sounded quite ok, but PRO sounded awful. There seemed to be added reverb jumping around in the background (some kind of stereo processing?). Also, sustained notes tended to vibrato a bit (these effects at 91% playing speed). Again, this was only with PRO mode.
> You missed my point. My point is that the worst case is when > you are performaing downsampling and the source signal has > significant amounts of audio energy about the nyquist frequency > of the destination sample rate.
In this case, that energy will be filtered out to prevent aliasing. Re: Procrustes "Pro" versus "Fast": For me, "Pro" sounds worse than "Fast" (i.e. a modulation appears) for tempos about 70% or lower, not 90%, even for the piano. And while it is true that the usable stretch range of "Pro" is narrower, its processes the transients much better. Conclusion: "Pro" is more suitable for music where there are a lot of fast transients, but the required stretch ratio is not so high. "Fast" is more suitable, if the required stretch ratio is higher, but there are not many fast transients. Of course, it's also almost twice as fast than "Pro", too. Regards, Audio DSP team http://www.dspteam.com/
Andor wrote:
> DSP team wrote: >>> I can't remember which exact sonata it was that sounded awful, but it >>> was Mitsuko Uchida playing one of Mozart's. I have the whole set of >>> Mozart Sonatas on my notebook, losslessly ripped from my set of CDs. I >>> picked a few odd files at random from those, and from a set of Mozart >>> concertos, to try. The really bad solo piano pieces were medium speed, >>> and not desperately complex. >> All right, I went and downloaded all available 30-second MP3 samples at a >> New Zealand site offering Mitsuko Uchida's performance of Mozart's piano >> sonatas (http://www.opuscds.com/cd/29490) and played them all at 75% >> tempo. Again, I didn't notice anything bad. > > There is quite a difference between the Fast and PRO modes. For piano > (Erik Satie, Gnossienne), the Fast mode sounded quite ok, but PRO > sounded awful. There seemed to be added reverb jumping around in the > background (some kind of stereo processing?). Also, sustained notes > tended to vibrato a bit (these effects at 91% playing speed). Again, > this was only with PRO mode. >
Interesting. I was using the PRO mode before, since that's supposed to be the higher quality option. I didn't even try the fast mode before. Now I've tried it, the fast mode is, indeed *dramatically* better. The results with that are pretty decent on the pieces I have been trying. Steve
>Interesting. I was using the PRO mode before, since that's supposed to >be the higher quality option. I didn't even try the fast mode before. >Now I've tried it, the fast mode is, indeed *dramatically* better. The >results with that are pretty decent on the pieces I have been trying.
Perhaps we should rename the "Pro" version to "Transient" or "Attack"?