Resynthesis questions

[cixelsyd]

Try alchemy or a neuron vs

[jdsowa]

When you resynthesize a sample what you are doing is picking 240 (or whatever) of the most important frequency bands that the sample shows on a graphic equalizer/frequency spectrum analyzer and you're creating a sine wave at each of those points. The reason a sine wave is used is because a sine wave is the most basic element of sound. Just a fundamental tone with no overtones. You can't do resynthesis with any other waveform or a sample because they have an overtone series, and the whole purpose of resynthesis is just to insert these thin little sine waves at these 200+ points, and it all comes together to make something that distantly resembles the original sample. If you wanted something that sounded like halfway between a bell and a dog's bark, I would just do resynthesis those two individual samples separately and then layer them on top of each other.

[codestar]

Yes I understand how resythnesis works I just work like to be able to insert a sample in the place of a sine wave for partials. I was thinking about being able to create voicing of horns made from something completely different. Layering by itself wouldn't work because I need each sample to have an envelope that matches the partials amplitude in order to achieve this effect

[Praxisaxis]

Originally Posted by cixelsyd

Try alchemy or a neuron vs

+ 1

Definitely give Alchemy a go... it kinda does what you're describing (though it does use either additive or spectral synthesis)

[codestar]

I took a look at alchemy. Very cool. I think it gets me 90% there, but I still need to be able I replace (at least a few oscillators) with samples. I wonder if I email them (and buy the license) if they ll give me access to the code to customize that function

[codestar]

Synth maker might be the way I have to go one this one

[polybonk]

Man I wish the Neuron didn't die.

[cixelsyd]

Originally Posted by polybonk

Man I wish the Neuron didn't die.

My neuron VS is alive and kicking. I run it on a Muse Receptor.

[Yoozer]

Originally Posted by codestar

Yes I understand how resythnesis works I just work like to be able to insert a sample in the place of a sine wave for partials.

But according to Fourier every sample you're going to insert also consists of sinewaves. I do get your intent though; your operation would create a series of spectral copies, based on the Fourier analysis of the sample itself.

I think what you want is slowly getting into deep Reaktor or Kyma territory. Question is whether it sounds anywhere decent.

Originally Posted by cixelsyd

My neuron VS is alive and kicking. I run it on a Muse Receptor.

Doesn't it slowly run out of memory when you switch presets? Because that's what it did when I owned one.

[cixelsyd]

Originally Posted by Yoozer

Doesn't it slowly run out of memory when you switch presets? Because that's what it did when I owned one.

I can confirm this is an 'undocumented feature'. Having said that I don't get the issue too often. I generally go in a program a single sound rather than sound shop through the programs.

I bought my neuronVS on ebay and it was the boxed version complete with the nuke and all the disks. The guy I bought it from also included a version that didn't require the nuke. I'm not sure who this guy is, but I'm guessing he was a beta tester. I've never heard of cracked NeuronVS's. Anyway when my G5 mac karked I had a go at installing it on the Receptor 1 revision C. Man I was amazed I got it on and working. Anyway, I guess I'll have it for as long as the muse holds out. I'm hoping to get an eventide H3000 VST for the muse as well. C'mon eventide where is it.

Anyway, I reckon there is nothing like a neuron VS. Kyma can do good stuff but I haven't heard anything sound like the neuron. That said I find it probably is best suited to scapes. ymmv

[gremlin moon]

@op It seems like you may be talking about convolution/cross-synthesis:
Music and Computers
SoundHack: Edit to Convolution
Convolution - Electronic Music Wiki

Synths that enable it:
Metasynth MetaSynth 5
iZotope Iris | OVERVIEW
Galaxy X Best Service - Best Service - Galaxy X - EN
SoundHack 0.872 User's Manual

I also think that one can cut and paste spectral resynthesis data in Alchemy. I would ask on their forum about the best workflow for it on KVR since it seems like they are pretty nerdy guys willing to go way in-depth about that stuff. Also there is someone from Best Service who posts in the Galaxy X thread on KVR and he goes into the whole process of working with Yellow Tools to create a convolution engine. I am not sure if the standalone Yellow Tools sampler has that function itself though.

Omnisphere has a lot of "CMT" samples -- which is what they call their "proprietary" morphing of sounds -- they are heavy Mac guys so I assume they use metasynth.

[codestar]

I think the difference is that I don't want to cut and paste or combine information I simply want to substitute an oscillator for a sample and use that partials amplitude and duration to act as an envelope for the sample. I found the composition that does this. Mortuos Plango, vivos voco by Jonathan Harvey. At first it will just be Bella and a boy singing. I trust you ll figure out when the effect I describe is happening. I don't want to resynthesize the samples that I want to place in the partials either I want them to contain all their harmonic content

[gremlin moon]

Vivos Voco was recorded at IRCAM (IRCAM - Wikipedia, the free encyclopedia). They have a commercial partnership with FLUX that markets their research and turns it into products. The product that contains the same DNA used in the John Harvey piece is called Trax. Flux:: sound and picture development
IRCAM-tools

Pricey products lol.

You could always ask Curtis Roads: http://clang.mat.ucsb.edu/biography_and_contact.html

[bluegreengold]

In the Harvey piece, the composer analyzed the sound of a large bell. For each sinusoidal component in the resynthesis, the composer substituted the sound of a sampled boy's voice at the appropriate frequency. The voice samples followed the analyzed frequency and amplitude control functions of the chiming bells, creating an eerie effect of a boy/bell chorus.

page_146

I wouldn't try to use a VST for this kinda thing, whatever the design of the plugin is, it's going to limit you in how you approach the sounds in ways that I'd consider unsuitable for really diving in and experimenting or doing detail oriented manipulations.

A combination of Csound/RTcmix would be my approach. The software is free, and learning to use it will be greatly edifying in itself.

[Denis Goekdag]

I think the only options are Kyma or Metasynth. Use Kyma if you need real-time, Metasynth if you don't have the $5k ;-)

In MS, you analyze the spectrum of a sample, then use a sampler as engine instead of a sine-wave. It's not as high-res as kyma because only the FFT amplitude data is used instead of Amp & Phase, so the pure resynthesis always sounds somewhat synthetic....but it gets you there.

[bluegreengold]

Originally Posted by Denis Goekdag

I think the only options are Kyma or Metasynth. Use Kyma if you need real-time, Metasynth if you don't have the $5k ;-)

The original inspirational piece never used anything that fancy. All that code that went into developing the Music V ancestors is available for free. A modern computer can do all the processing without breaking a sweat. "Real time" phase vocoding is going to put serious limitations on what you can do. This is a conceptualize, compose and program(code) type of technique.

[codestar]

Wow great stuff guys thanks!

[polybonk]

Harvey: "Vivos Voco" - YouTube

After listening to that track you are talking about I think you would love both Kyma and the neuron VS.

I assume that you are coming at it from a musical perspective and not a theoretical one yes?

[Denis Goekdag]

Originally Posted by bluegreengold

The original inspirational piece never used anything that fancy. All that code that went into developing the Music V ancestors is available for free. A modern computer can do all the processing without breaking a sweat. "Real time" phase vocoding is going to put serious limitations on what you can do. This is a conceptualize, compose and program(code) type of technique.

I was replying to the original question of using samples instead of sine oscillators for resynthesizing partials after a spectral analysis. I wasn't implying that that method would be required to create similar results to that piece.

[bluegreengold]

Originally Posted by Denis Goekdag

I was replying to the original question of using samples instead of sine oscillators for resynthesizing partials after a spectral analysis. I wasn't implying that that method would be required to create similar results to that piece.

Yeah but you did suggest that "the only options" were expensive hardware or software, however there are numerous free software options available that can do this. PD, rtcmix, csound just to name a few.

[Denis Goekdag]

Well, if you want to deal with that type of environment, yes, you 're right, these are options. So is Max/MSP with Jitter or CDP if you can spare the time and energy. In *my* book, these are all very interesting, but workflow-wise they end up eating lots of time more often than not (actually, even Kyma is "borderline" in that regard IMHO).

YMMV.

-d

[Sinner]

I'm much more a string instrument player so I could be off on my synthesis terminology but I got this just because they were having a sale and the demos were nice...it lets you load your own wavs or use oscs.


De La Mancha
"Scylla is a VST instrument capable of sample-playback combined with subtractive synthesis to produce a wide variety of sonic possibilities ranging from lush pads to far-out soundscapes and effects via scorching leads and chunky bass. Scylla can be used for most any application and is only limited by the user’s imagination.

Scylla includes three different types of oscillators that can be modified to a wide variety of styles. Three built-in samplers can accommodate .wav audio samples, which can be played back in grain table, waveform, sample or single hit format. All the audio can be modified and modulated through LFO, volume and filter envelopes, Ring-Modulation, and Unison."

de la Mancha Scylla

[foodeater]

Cool song!

To me the track sounds not so much as it's using a sample in place of sines as it's using the re-synthesis of the sample!

Metasynth is an easy way to get this type of sound although I would approach it in a different manner (with no convolution needed per se) unless you want to explore that same process. If you like playing with this kind of stuff it's a great tool!

Logic (and main stage) has a plugin called EVOC 20 track oscillator that lets you use another audio source.

I also sometimes do something similar to this with my korg radias and the vocoder, you can only record 8 or so seconds and you can just use an external in, but you have to do some fancy programming to get interesting blending. Nice to stick Very fast and slowly modulating wave shapers in various places.