This page is available in English only

Sparky

Sampling Spectral Resynth/Real Time Sound Mangler

(1.1K Votes)
2.1 (Updated 16 years ago)
884.4kB
June 10, 2004
Reaktor 5 or lower

DESCRIPTION

I can't decided whether this is best described as an effect processor or a synth. It can be used as a spectral sampler, a vocoder, a formant shifting synth, a strange evolving soundscape maker, a spectral delay - lots of things. All the sections and controls have tool tips. Most of these make sense.

* Very important - midi clock must be running for anything to work *

Basic idea is that sounds are split into frequency bands. The number of voices control the number of bands. Since event tables must be fixed size max voices is 256. Due to careful use of event processing and polyphonic tricks 256 voices run with no problems on my 867MHz Powerbook. Although you'll see you don't always want this many bands.

There is an analyser which can be used to grab sounds and store them in the table. Or can be used in real time for spectral delay/resynth/vocoding type stuff.

The table can be scanned and shifted in various wibbly ways. (There is a six stage, three dimensional looping envelope.) It can feedback on itself, and can be excited with a spectral burst.

The output of table drives the resynthesiser which can be additive, or a bank of filters (like a vocoder).

Consider the snapshots to be starting points for further investigation. The sounds produced will vary greatly depending on the number of voices and the contents of the table.

As a synthesiser it is monophonic - it has to be this way because I'm using polyphony to model the partials within a sound. If Reaktor had some sort of vector construct I would love to make a polyphonic version.

Deep down it's based on SolarX's Metasyn64 v0.1, but really it's a complete rewrite. Sparky uses about 1/4 of the CPU of Metasyn64.

I won't be adding any extra features to this ensemble. I'd like to change the way the table is addressed, and the flow of events. I also want to try writing a DFT algorithm. All of this implies a complete rewrite. I don't know when this will happen.

There are a couple of known bugs: jumpy behaviour when turning various bits on/off; free running envelopes don't start when changing snapshots; sometimes loop points can be missed; there is a small amount of weirdness here and there.

I've included a small select of table files with the ensemble. You can load and store stuff in the usual way.

***

Spoonbung want instant gratification, Spoonbung get instant gratification. New snaps added. For simple morphing use snap 34 "SM - Morph Record". Midi note or pressing to Step button advances step. For playback use snap 35 "PL - Manual Morph (Use Y Slider)". You guessed it - use the Y slider to morph between spectral samples. Or, use snap 36 -"PL - Looping Morph" to loop continually between two spectral samples. Edit the ScanEnv step positions for different effects. I suggest sampling the phonems ah ss ho le for fun and laughter.

COMMENTS  (45)

Marcelo k
4 years ago
Thanks for sharing!
Christian Findlay
10 years ago
This looks really interesting but for the life of me, I can't figure out how to use it. Is there are help document/tutorial somewhere? Am I supposed to route audio to a track with this thing on it? Is that how it works?
Molec
10 years ago
This is brilliant. I'm going to spend some time figuring this out properly because it's just so different.
Grant Miller
11 years ago
I intend to use this in my music. Thanks for a great contribution to the library.
Jonas Broberg
11 years ago
Fantastic! Thank you very much!
Maximedes
12 years ago
hey ho! thats what im talking about! very useful! thx
Robin Barklis
13 years ago
Excellent. Thank you.
Christopher Soulos
13 years ago
Wha the f...! WoW! Love the noise makers.
Ulrich Reuter
14 years ago
bravo, bravo!
michael huckaby
14 years ago
incredible. i havent gotten my head around the whole thing, but its amazing.
Norman Wilson
15 years ago
Sorry I haven't been on this site for a long time. I wrote it in R4 so it definitely doesn't use any R5 features. NI must have converted it.
Sven Röhrig
16 years ago
found it :-), sorry
Sven Röhrig
16 years ago
31. I just wondered if its possible with Reaktor to do this kind of sonograms and you did it.... Wow... That is really somethig... Ups... Reaktor 5...Ok... I vote anyway just what i see is enough to give it a top vote allready... Is this possible with R4 aswell? anyway regards Sven
oscar finch
16 years ago
Interesting concept but still getting my head around it :). Great GUI and capable of some very intriguing textures none the less. Makes for some interesting emlodic backing with SineBeats I might add. Thanks - great work.
Dave Bourke
16 years ago
I got sound out of about half a dozen of the snaps - the rest generated only silence (Mac). Looked interesting though...
Aaron Chutz
16 years ago
Got a little excited there. Great synth that makes otherworldly sounds. Highly recommended for anyone into IDM, Noise, Sound Design and the like. Everyone should have this!!!
Aaron Chutz
16 years ago
OUT-FU@%ING-STANDING!!!!
Miguel Santos
16 years ago
I´ll try it out now
Omar Sharan
16 years ago
This thing looks cool . gotta try it
George Gyulatyan
16 years ago
Interesting beast. I'll be damned if I can figure out how/what you've done or what does what, but all I can say is that it does some crazy stuff especially when working it interactively with feedback from my mic/monitors as Sparky outputs. Sigh... how do guys like you get to that level of thinking? My respects to you. Gets a 10 from me!
Carl Taylor
16 years ago
this is crazy, I dont quite understand it, but im getting some cool sound out of it anyway..
Norman Wilson
17 years ago
I haven't checked this recently but... The SM snaps are for sampling. They are provided as example setups for capturing spectral samples into the big table. They shouldn't produce any sounds on there own. The PL snap are for playing back samples. If the table is empty you won't hear anything. PL snapshots will not change the contents of the table. Any other snaps will. (So probably this is not a bug in 4.1.3 - on the other hand it might be). So the usage is something like: select a SM snap. Make a recording (usually triggered by key press or pressing the trigger button in the lower left corner). Select a PL snap. Have fun. This was actually what Sparky was intended to do first and formost. All the other stuff evolved later. Unfortunately because of the nature of event tables it's hard to make this process fool proof - you really have to record your own. If you do take the time you can get some absolutely nutty results.
devin fleenor
17 years ago
amazing ensemble - one of my absolute favorites. I can't get much volume out of snapshot 22 - chromatic 100 band. Also, I get no sound whatsoever out of many of the SM & PL snaps. I really hope this version isn't broken because of 4.13 on OSX.
Rüdiger Ehrmanntraut
17 years ago
cold synth with strange voice modulations, becomes my favourite playable melodic noise-maker; thanks for this high-Q thing/ensemble
Dragan Jovanovic
17 years ago
HIGH there!!!! Thiz thang iz WOW! Great for soundscapes and stuff. I can´t get enough of this sort of .ens´s You´ve spent so much work on that and we just download it...I love the REAKTor Community!
Norman Wilson
17 years ago
The next delite is simulataneous analysis and resynthesis.
David Dayneko
17 years ago
Awesome! I finally figured out how to record the data tables myself, great for running vocals in and getting cool ambient textures out.
David Morris
17 years ago
Outstanding. This is one of those pieces of software that takes a lot of study but is well worth it. Many nice interface ideas too. Thanks--especially for a well annotated instrument.
Roman Belavkin
17 years ago
metasyn64 uses polyphonic tables, you probably looked in an old version. the only cpu saver is indeed the panner/spreader, which can be quite a lot actually
Norman Wilson
17 years ago
Getting good sounds using bang, table feedback, and the scanning env is a bit of a black art but I definitely recommend experimenting with these features.
Robert Croft
17 years ago
this is really interesting. it does have a learning curve to it. fantastic work though.
Norman Wilson
17 years ago
Actually I've just looked at Metasyn64 v2 - it does do spreading. The original version didn't. BTW Sparky's spreader in sym or asym does put each voice into a different position in the stereo field.
Norman Wilson
17 years ago
Thanks for the comments. About CPU usage. The biggest saving in CPU for Sparky over Metasyn64 is the way I scan the table. Look inside Metasyn64. His table is monophonic, he scans it both horizontally (in time) and vertically (in frequency) with an iterator module then directs the values to each of the 64 voices. This is totally unnecessary. In Sparky the table is polyphonic. Each polyphonic voice looks at a single row (kind of) so the vertical scanning is implicit. This uses about 1/4 of the CPU as Metasyn64. Other savings come from the clock architecture which makes sure no unecessary events are generated. As for the stereo spreading - I didn't think Metasyn64 did this at all.
Simon Binks
17 years ago
voting
John Williston
17 years ago
Very wacky, but in a good way.
Roman Belavkin
17 years ago
nice work! i thought of uploading another version of metasyn64 with different scales and resynth methods (e.g. filtered noise), but now looks like it is not necessary. btw, i believe the cpu% is lower because of the stereo output macro works in a different way. it saves cpu, but does it really do the same job positioning individual voices?
Wolfgang Webermann
17 years ago
Loading data into table doesn't work on my pc... a windows-problem? Phantastic instrument by the way! Vote 10! We want more!
Norman Wilson
17 years ago
How to load data into the table: Ctrl click on the table to get a menu up (on a dos box this is prbably right click). Go to file->load table data. The files I posted should all load - they certainly work on a Mac. Note these are Reaktor table data files, not samples as such (wav etc). To make you own recording you have to use the analyser. All the files I posted were recorded in Reaktor stand alone mode using the built in microphone on my notebook. There is a snapshot - I think it's call SM - key trig sample (or something). After you've recorded switch to one of the playback snapshots - PL - Hifi I think. You have to make sure the output transposition matched the input tuning. Getting good results out of the analysiser requires a bit of tweaking. I will write some sort of manual but I can't promise when it will be.
liv
17 years ago
Very interesting contribution
GG.G SakabeaT
17 years ago
Fantasic,Magic,Hypnotic !!!you did only one mistake!!you put it in the Additive categorie,you'd better put it in the Addictive one....;-)
Norman Wilson
17 years ago
Mr Hopper has highlighted a number of issues. The quick solution to CPU overload is to use less voices. 256 is the maximum but you don't really need that many. 64 is fine for most spectral stuff. (Although 256 sounds quite creamy). For vocoderish stuff you might want to use only 10 voices. Point 2 - some of the snaps will overwrite the table contents, some of the snaps will depend on having something in the table. This is a problem with Reaktor not storing table contents with snaps. You can reload the table in the usual way - ctrl-click->file->load data... I've tried to categories the snaps - NZ - evolving noises - sometimes key triggered. SM - sample/analyse (resynth turned off). PL - Straightish playback. SY - Pure synthesis. FX - Mash up the input in real time. VC - Vocoder - processes two inputs - you may have to mess around with some connections to get this to work. Out of these categories the only one that doesn't over write the table is PL. One of the reason I took so long to post this (it's been in developement for months) is that I kept finding new uses for it - and thus extra features. It's really pretty deep and I could do with writing a manual for it. BTW you might want to try reloading the ensemble and selecting snap 7 to find the secret message.
Kevin Hopper
17 years ago
Very, very interesting creation. Snap 24 ( the one formanty that opens when you open the ens.) is absolutely facinating. However, as soon as I get to the other snaps a 'processor warning' comes up. So I switch to a lower bandwidth and go to the snaps again...but this time they are either silent or make the odd high frequency noise. Lowing the amount of voices seems to turn what's barely there into sine tones. Doh!I blame my poor underpowered Mac 665.
raphael esterson
17 years ago
This is really fantastic. They said it couldn't be done... not just a thorough implementation of a new technique, - a very creatively responsive one. Thanks!
Phil Durrant
17 years ago
very interesting, and an important upload. it is quite a head scratch but for this one, i am glad to make the effort
Markus Brandt
17 years ago
very interesting fx/synth ;) - i like the idea behind this little monster. hey and it soundsa nice to - very cool vocoder sounds. best regards
now