KVR Audio

WolfKay · Post by **WolfKay** » Fri Dec 16, 2016 8:37 pm

That's beautiful!

Keep 'em coming!

Xenakios · Post by **Xenakios** » Fri Dec 16, 2016 9:26 pm

Another video, demonstrating reversed plugin processing :

ZebraTamer · Post by **ZebraTamer** » Sat Dec 17, 2016 10:17 am

WolfKay wrote:That's beautiful!
Keep 'em coming!

1+ super awesome!

CinningBao · Post by **CinningBao** » Sat Dec 17, 2016 1:34 pm

Yeah, man, this is incredible stuff - lovely sequencer flow

Are the CDP guys aware of this at all? I think they'd be mildly impressed..?

Xenakios · Post by **Xenakios** » Sat Dec 17, 2016 2:25 pm

Video demonstrating some of the CDP programs, finally appropriate for this forum thread.

Xenakios · Post by **Xenakios** » Sat Dec 17, 2016 2:41 pm

CinningBao wrote: Are the CDP guys aware of this at all? I think they'd be mildly impressed..?

Maybe, I've posted about this in the CDP forum.

marmess · Post by **marmess** » Sun Dec 18, 2016 5:48 pm

Ouf...your are building a cathedral here!
Cheers!

Xenakios · Post by **Xenakios** » Wed Dec 21, 2016 11:23 am

Probably of interest only for coders.

Lua scripting in the sound Montage node (previously named SequencerMixer).

I should probably start another forum thread about λ...

klovn · Post by **klovn** » Fri Dec 30, 2016 11:11 pm

Xenakios · Post by **Xenakios** » Sat Dec 31, 2016 1:29 am

Release clearly not going to happen this year.

keyman_sam · Post by **keyman_sam** » Sat Dec 31, 2016 2:45 am

Xenakios wrote:Release clearly not going to happen this year.

We can wait, but please make it fast!

So to summarize - this is basically a front end for ALL CDP nodes, right? i.e. all the modules available in CDP will be available here with the same parameters?

BTW this whole insta-rendering thing that's happening here is EPIC.

Xenakios · Post by **Xenakios** » Sat Dec 31, 2016 3:36 am

keyman_sam wrote:this is basically a front end for ALL CDP nodes, right? i.e. all the modules available in CDP will be available here with the same parameters?

All the ones the Renoise CDP front end developers added in their Lua code. So almost all of the CDP programs and their operating modes, with some of the multichannel/surround ones missing/commented out because Renoise has no support at all for audio files of more than 2 channels. (I need to sort those out myself at some point. I've already added a few of the multichannel programs.)

The CDP stuff comes with the caveat that many of the programs won't run with the default parameter settings or don't have sensible parameter ranges. (As great a job Djeroek did for the Renoise front end Lua script, there are missing default parameter values and/or parameter ranges that are not sensible.) But that is something that will just have to be gradually sorted out. (I might be able to come up with some kind of an automated system to do some basic checking on which of the programs run correctly with their default settings and which don't...That might make for an amusing video too, if I make the automated test update the GUI while it's doing the test.

edit : I started coding the automated test, but it needs to be done carefully so that I can abort it if it takes a really long time, it starts filling the hard drive too much etc...)

cron · Post by **cron** » Sat Dec 31, 2016 4:12 pm

Xenakios wrote:edit : I started coding the automated test, but it needs to be done carefully so that I can abort it if it takes a really long time, it starts filling the hard drive too much etc...)

Regarding filling the hard drive too much, I imagine the spectral processes could make this a nightmare. CDP's .ana files are huuuuuge compared to, say, .spdif files from SPEAR. They're not even particularly compressible with tools like 7zip or rar. I was really surprised by this as I thought there'd surely be tons of dead space in there given the size.

It looks as though doing a spectral analysis with, say, a window size of 8192 produces a file of equivalent size to an 8192 channel .wav. Lord knows how users effectively dealt with files of such size in the early days of CDP.

Anyway, these videos look absolutely incredible. This thing looks more ambitious than I'd ever dared imagine. Thank you.

Xenakios · Post by **Xenakios** » Sat Dec 31, 2016 6:34 pm

Test results so far, for a 1 second mono wav-file as the input :

Finished test for 221 PCM and 106 pvoc nodes. Elapsed time 21511.990747 milliseconds. 117 errors. 42 parameter related errors.

276 files were produced, taking 158 MB of disk space.

So not too bad, if not that great either with the errors. In theory most of the default parameter errors can be fixed by writing other values as the defaults. The remaining errors are crashes and miscellaneous errors. (Fortunately the crashes don't affect λ itself.)

The timing seems to indicate on average the CDP programs will run at least 10x real time speed on a single CPU core. (So would be feasible to eventually make the CDP programs run in real time with immediately audible parameter changes etc...)

edit : More timing info (note that the elapsed time is for the total of all of the CDP programs that will run, not just a single one) :

So looks like there's on average some overhead for short input files. (Which would be expected since new files need be created, new processes launched and so on. But there might be some room for optimization in the λ code too.)

Some weird behavior which might be because of algorithmic inefficiencies or a mistake in the benchmark calculations :

But if this is correct, it would seem the average real time processing speed could be much greater than the about 10x mentioned above...

Jedinhopy · Post by **Jedinhopy** » Sat Dec 31, 2016 7:20 pm

cron wrote:spdif files from SPEAR.

They are not called spdif.
They are called (.sdif) .

Composers Desktop Project, still relevant ??