Timestreching Engine

Though I don´t really feel the need to have this feature, some other people might, there where some discussion about it in another topic but I just wanted to enlighten the developers that there is a timestrecher engine they might be able to use…

http://sky.prohosting.com/oparviai/soundtouch/

9xx allso do a trick

Engine looks pretty cool… Assembler code included for x86… Do not know what LGPL license is though… (didn’t feel like reading :)) But is it less restricting so you can make changes without giving back the code?

On the side of the page it says:

Do you like SoundTouch? Are you using SoundTouch library for a commercial application?

Then you may wish to support the development effort by making a donation!

Cool… I’m curious what the developers think about this…

The answer is no. LGPL is mostly used for libraries in order to permit linking those libraries into non-free programs. But that doesn’t mean that you can do whatever you want with it. The code of the library, either changed or original, must be redistributed with the binaries. But it means that a non-free software (for example Renoise) can make use of this library as long as it is clearly stated and that all parts of the code that is based on the library is redistributed with the binaries.

If you want to know what free means when it comes to GPL then read this
http://www.gnu.org/copyleft/gpl.html

;)

A GPL program can cost money, but the source code has to be completely open.

You shouldn’t. What he meant was that if anyone wants to participate,
they should just tell us. We’re not a completely closed project.
Quotes out of context are not to be trusted…

Judging from the small response of qualified people with time to spare
(in particular the latter…), it seems to me like a waste of efforts to make
Renoise opensource. Extra work, little gain.

agree. An API to let people code effects, generators and meta devices would be much better

… And as we’ve said many times now, if anyone is burning for some feature,
and know how to program, they’re welcome to join us and implement it.

VST(i) should work fine, really.

User coded metadevices… shivering

Anyway, none of these would make sense to do now, since the ways
Renoise handle this and that is likely to change a lot over the next year.

Maybe you could post a wanted ad on the main page on the website, this might get some more attention from accidental visitors. Also the page gets indexed by google or other crawlers… People can also then refer to the page when they speak to someone that might be intrested in programming for Renoise… Just an idea… (sorry offtopic I know)

Are there any VST’s for timestretching with the pitch being retained? It could be a useful tool for transcription, but then for it to be useful, the audio would have to correlate to the position.
I.e. not just be triggered by Renoise, but actually be there, in the flesh, so to speak, enabling you to hear the audio played from a certain sample position, based on it’s correlation to the pattern resolution, and when it was triggered.
Am I making sense? Has this been suggested before? Is it possible?

timestretching is something which cant be done “good” at this point without some serious cpu-power. all those realtime-timestretching algorithms I know of sound baaaad (NI intakt for example) and the only one which sounds good (timefactory) is takes about ten times the length of the sample to compute the timestretched version (I have a 2,6 ghz athlon).

ableton live has a good time stretching enine, and it doesn’t take very long for it to calculate the new sound.

yes, but it won’t be a problem if you timestretch then save your wave. unless you timestretch all the samples in realtime, you might have a problem then.

anyway timestretching is a lot of interpolation, which isn’t very intensive mathematically or computationally ( O(n) i suppose) - but we’re talking about really high bit rates, so practically speaking there’s a lot of calculations.

and sometimes interpolation is not enough, and you need to throw in other DSP in it… so it all adds up.

If you have a timestretching algorithm which is based solely on interpolation,
and works in O(n) with respect to the number of samples, you better patent
it right away, and you’ll be rich in no time. :rolleyes:

hehe… maybe i was crappy :P

anyway i didn’t really search for any algorithm, just giving my 2 cents (forgot to disclaim it). maybe it’d be good if anyone can post a few snippets of info so i can brush up on DSP… :unsure:

A good general advice is to not give your 2 cents when you’re not sure
you really have them… If you don’t know any timestreching algorithms,
then don’t claim you do. Simple as that.

There’s a link to some kind of “formant based” timestretching in this forum,
in a post from one-two weeks ago. Search yourself, I’m too lazy.