voiceofsnow

Talk about all things "KVR Developer Challenge" related.
xoxos
Mr Entertainment
12018 posts since 30 Apr, 2002 from i might peeramid

Post Thu Dec 08, 2016 8:51 am

so far the only entry that didn't crash or denormal on me, so good there.

most people won't notice but this is the first freeware vst process to avail this technology, so i appreciate that. i've been intending to implement lpc or cepstral for (re)synthesis for some time. from what i know about cepstral processing, source-filter separation would use some extra work, i found the separation here worked well for the application of creative signal shaping, the note about complex/noisy signals of course hinting at one of the exploitable regions, if you're that kind of composer ;)

thank you for sharing!
you come and go, you come and go. amitabha xoxos.net free vst. neither a follower nor a leader be
tagore "where roads are made i lose my way"
where there is certainty, consideration is absent.

DSmolken
KVRian
1380 posts since 20 Sep, 2013 from Poland

Re: voiceofsnow

Post Fri Dec 09, 2016 1:44 am

I tried to get it to separate the pitched signal of a bowed double bass from the bow noise. The "noise" is still clearly pitched but much thinner and noisier - not as completely separated as I might like, but still has all sort of interesting potential for turning instrument samples into more flexible models. Need to see how it will do on rough vocals, pizzicato...

Eleventh
KVRist
325 posts since 1 Dec, 2012

Re: voiceofsnow

Post Fri Dec 09, 2016 3:52 am

I think this is a hidden gem among the entries, and one of my favourites definitely. During initial testing I found out that it works _really_ well for de-harshing fizzy amp sim guitars.

One thing I wish the plugin had, is a tilt filter also for the tonal part. Could be useful when you decrease the noise, and then tilt the tonal brightness back up.

Anyway, excellent little plugin, so many thanks to the dev! :tu:

User avatar
Aloysius
KVRAF
22274 posts since 11 Aug, 2008 from a computer

Re: voiceofsnow

Post Fri Dec 09, 2016 4:50 am

i 4gt 2 chek DIS 1 out. U nevr know, it mA mAk it in2 my fInL 5.
This little piggy went to the butchers ...

xoxos
Mr Entertainment
12018 posts since 30 Apr, 2002 from i might peeramid

Re: voiceofsnow

Post Fri Dec 09, 2016 5:28 am

DSmolken wrote:I tried to get it to separate the pitched signal of a bowed double bass from the bow noise. The "noise" is still clearly pitched but much thinner and noisier - not as completely separated as I might like, but still has all sort of interesting potential for turning instrument samples into more flexible models. Need to see how it will do on rough vocals, pizzicato...
i may well not have a clue what i'm talking about here,

afaia, this is performed by a cepstral process, which is basically a "spectrum of a spectrum" - at that point, the cepstral data has to be split to separate source ("residue" iirc) and filter data.

as you may imagine, writing an algorithm for this to adapt to any input is likely to be a bit intensive, from what i know about it, i'd intended to leave this as a user setting for resynthesis. it sounds like there's a fade or something here, so as said, not suited for rigorous analysis, but enough to be interesting and creative.

my perspective here is from listening to other source-filter separations (those with the STK) which are much more "separate".

i had a noisy field voice recording which it seemed to make much more "listenable," like hair conditioner for your audio :p bouncy, silky, manageable. cpu is about right on for what i "understand" it's doing (two ffts and a pow on each bin magnitude! plus the separation algorithm and ifft).
you come and go, you come and go. amitabha xoxos.net free vst. neither a follower nor a leader be
tagore "where roads are made i lose my way"
where there is certainty, consideration is absent.

sundevstudio
KVRer
1 posts since 30 Nov, 2016

Re: voiceofsnow

Post Sun Dec 11, 2016 5:17 am

Eleventh wrote:I think this is a hidden gem among the entries, and one of my favourites definitely. During initial testing I found out that it works _really_ well for de-harshing fizzy amp sim guitars.

One thing I wish the plugin had, is a tilt filter also for the tonal part. Could be useful when you decrease the noise, and then tilt the tonal brightness back up.

Anyway, excellent little plugin, so many thanks to the dev! :tu:
Guys, thank you for your responses!

I've update version to 1.0.1. Now it includes Tone_Tilt parameter as Eleventh asked. Also I added "Separation_Slope" parameter it can help you to brighten up tone signal when you remove noise component. By default (in 0 position) the plugin uses pink-noise like slope for separation process. It gives better results for me but tone component becomes darker. In 1 position this slope is completely removed so noise component becomes more white-noise like. It may works better in some situations.

2xoxos: Cepstral processing is used indeed, but to construct dynamic min-phase filter used to split both components. For my experience min-phase filtering gives better results here.

Eleventh
KVRist
325 posts since 1 Dec, 2012

Re: voiceofsnow

Post Sun Dec 11, 2016 7:13 am

Amazing, thanks for the update!

The Separation Slope parameter works well for brightening the tone part :tu:

Hmm just had an idea... could this process work dynamically? Especially in the case of de-noising saturated guitar sounds, it would be desirable to let the initial attack through the noise reduction, but I'm sure it could be useful on other sources too. Maybe something for future development...

Anyway, thanks again for this great plugin. I'm sure we'll see this tech utilized in lots of future plugin offerings.

xoxos
Mr Entertainment
12018 posts since 30 Apr, 2002 from i might peeramid

Re: voiceofsnow

Post Sun Dec 11, 2016 10:07 am

sundevstudio wrote:added "Separation_Slope" parameter
:o haven't tried it yet but it can only make a good thing better.

may i extend to you my deep, heartfelt gratitude for offering this process to vst users.
you come and go, you come and go. amitabha xoxos.net free vst. neither a follower nor a leader be
tagore "where roads are made i lose my way"
where there is certainty, consideration is absent.

User avatar
hate9
KVRist
30 posts since 31 May, 2016

Re: voiceofsnow

Post Mon Dec 12, 2016 7:14 am

I really wish there was some sort of option for sensitivity, but other than that, I could wish for absolutely nothing more. This is by far my favorite plugin from this year's dev challenge, and I sincerely hope it wins.

User avatar
DJErmac
KVRian
602 posts since 25 Apr, 2009 from France

Re: voiceofsnow

Post Thu Dec 15, 2016 4:56 pm

Most impressive plugin this year, five stars.

rcrath
KVRist
69 posts since 3 May, 2005 from Honolulu

Re: voiceofsnow

Post Sun Dec 18, 2016 4:06 pm

Ho sundevstudio! I really like this plugin and already have half a dozen use cases for it. I use an old pitch shifter, son-of-apitch by chrysalis, because it does polynotes on a guitar well and is low impact on processor. 1) filtering out the noise component before sending the guitar to sonofapitch makes the recognition better and makes it possible to do a really neat electirc 12 string knock off. 2) Adding just the noise component to a slightly crunchy fender amp brings out a sort of "throatiness" I associate with Vox amps, which is pretty cool as I've never been able to quite nail the "jangle" in S-Gear amp sim cuz no vox emulation. both of these require setting voiceofsnow as a send effect rather than an insert because of the one main issue holding this back. I know spectral processing is expensive, but the plugin adds a noticeable latency. That actually works for the above two items. In the twelve string it mimics the distance between the two strings in a pair, although the latency is more than the delay in actual playing a twelve string, it still works in smearing/delaying the attack of the octave strings a bit. It works as a sort of slow attack speed for the jangle amp. For use as an insert in recording I think it would need a latency compensation output. for use in live situations, the latency is too noticeable to use as an insert. Might make sense to offer a "less bands - worse analysis - lower latency setting or slider." I have always been interested in this re-analysis function and it is great to see it in a plugin. My fav so far!
my experimental improv duo, rreplay.
more of my music
occasional experimental sound blog all things vst, guitar synth, and more.
The Digital Guitarist.

xoxos
Mr Entertainment
12018 posts since 30 Apr, 2002 from i might peeramid

Re: voiceofsnow

Post Sun Dec 18, 2016 5:46 pm

u no wot -

copies/pastes from notes -

You can't set your plugin's latency in SynthEdit, however you can hack in in later.

Compile your dll out of SE and open it in a Hex-Editor.
Take a Hex-Editor (e.g. UltraEdit) and search for the string (not in ASCII code) -> C7 46 68 66 45 6F 4E <- then change the 68 to 50 and the following four bytes to 00 00 00 00.

C7 42 50 00 00 00 00 represents Latency 0 Samples. To calculate your correct settings just take the windows calculator, type in your latency in samples in dec-mode and switch to hex-mode. Then you have to add the 0 (zeros) so that you get four bytes.
example: 1024 samples are hex 00000400 (windows calculator shows only hex 400), 16 samples are hex 00000010. Now you have to order the bytes in Little Endian, so reverse them:
C7 42 50 10 00 00 00 represents Latency 16 Samples
C7 42 50 00 04 00 00 represents Latency 1024 Samples

After done so, open your PlugIn in Cubase and take a look at the plug-in informations, now you should see your new latency setting.
you come and go, you come and go. amitabha xoxos.net free vst. neither a follower nor a leader be
tagore "where roads are made i lose my way"
where there is certainty, consideration is absent.

Eleventh
KVRist
325 posts since 1 Dec, 2012

Re: voiceofsnow

Post Mon Dec 19, 2016 6:36 am

rcrath wrote:filtering out the noise component before sending the guitar to sonofapitch makes the recognition better
Hey that's a great idea! Gotta try if it improves detection on Jamorigin MIDI Guitar..

camsr
KVRAF
6877 posts since 17 Feb, 2005

Re: voiceofsnow

Post Fri Jan 13, 2017 6:40 pm

Real easy to use, thanks!
Image

Return to “KVR Developer Challenge 2018”