The ultimate Additive/Subtractive Universal Synth to replace all others?

[JSBach1685]

Hi guys,
Have been experimenting with Additive synthesis in custom C programs and Audacity for a very long time. The aim being the perfect replication of any sound. Comparing this to commercial software a few issues become obvious.

The best Fast Fourier Transform implementations are still no where near accurate enough for sound replication.

No additive synths take into account all characteristics of every harmonic in a complex evolving sound like an acoustic piano.

Other spectrum analysis techniques have been forgotten in favor of FFT. For example Wavelets, phase filters etc.

In my experience to perfectly imitate a grand piano c4 note you need at least 30 separate harmonics at any point of time.

They all have differing attack/sustain/decay profiles. They are not perfect sine waves (perfect sine waves don't exist in nature).

They are actually very slightly AM/FM modulated. They don't follow perfect frequency ratios. Etc etc etc...

5) There is no commercially available program that can scan a sound for its frequency content and recreate an IDENTICAL-sounding replica with the option of tweaking or deep analysis.



A mate said I should polish-up my program, Give it an intuitive user-interface and put it on steam to guarantee some sales before it's eventually cracked.

So is there even a market for this? would it sell better as a hardware synth or a software synth? could you market this to the scientific community (analyzing bird/bat calls, electronics) and the electronic music community at the same time?

Before you jump up and down exclaiming "have you tried program xyz" Trust me, I have spent way too much time trying out stuff from ImageLine, Native Instruments, Steinberg and small independent devs.

[Synthack]

.........

[JSBach1685]

Okay cheers!
Oh yes Urs Heckmann. What a legend.
Yeah, step one) make the finished user interface,
step two) Demonstrate the Synth on Youtube, comparing it to other synths, see what reaction you get.
Step 3) THEN start contacting the relevant people / organization.

Hadn't heard of Juice, Will check it out. Could be good for prototyping.
First I will code it all fully featured in C to demonstrate it as fast as it can possibly be on windows.
After feedback, It could then be ported to standalone, VST, AU, VSTi on Windows, Ubuntu, iOS.

There is a reddit discussion on the topic here

[xoxos]

at least thirty harmonics? that ought to strain the most robust fft implementation

but of course they can't do anything anyway

allegedly


holmes go easy on youself

[N__K]

A mate said I should polish-up my program, Give it an intuitive user-interface and put it on steam to guarantee some sales before it's eventually cracked.

So is there even a market for this? would it sell better as a hardware synth or a software synth? could you market this to the scientific community (analyzing bird/bat calls, electronics) and the electronic music community at the same time?

(emphasis mine)

I believe that most of synth-buying people don't understand the creative benefits of additive synthesis quite yet. Harmor made basic operation rather easy and still many users think it's the engine room of the Starship Enterprise.... which, from a certain point of view, it is.

So for mass popularity of additive synthesis, we probably need a world in which all synthesists understand the concept of sound as a sum of partials; that is, high school physics. To be blunt, it is a lot to ask of the world - but on a positive note, on every continent there is a growing number of people who get access to that level of education, and some of them will become computer and synthesizer users. So there is hope.

As for format, I suspect the right choice would be to realize it as a VST plugin and perhaps standalone software that can run on ubiquitous platforms: x86-64 and ARM; Windows, Android, iOS, macOS; also Linux if at all feasible. The widest audience is probably still on Windows x86-64, as cheap laptops are everywhere; but in five to ten years ARM-based alternatives may overtake it.

Also, both in near future and beyond, it is extremely important to realize a user interface which enables extensive control over partials and their changes in time while remaining relatively easy to use. In this regard, one can compare Image-Line's Harmor and Morphine: I believe that UI is a big part of why Harmor is more popular, although it also surpasses most of Morphine's functionality.

So I think that UI is probably a bigger challenge than the synthesis engine; and for success in VST market, the synth will probably have to have a "simplified page" comparable to Harmor's main panel or Vital's "Voice" page. Without that, I believe that large segment of potential users will think it's too complex for them.

Before you jump up and down exclaiming "have you tried program xyz" Trust me, I have spent way too much time trying out stuff from ImageLine, Native Instruments, Steinberg and small independent devs.

Sorry for a bit of jumping, but I must
I rely on additive synthesis daily, and my current tool of choice is Harmor.

In the ten years since its release I have not found a better combination of usability and capability in additive synthesizer. It excels at letting user make many different kinds of sounds (including inharmonic), and effectively surpasses everything that came before it.

The strength of Harmor is that it isn't only a "resynthesis rompler"; but that it's a creative synthesizer with user interface that does a very good (albeit imperfect) attempt at making everything easily controllable. I believe that that is the key to making additive synthesis commonplace: it has to be designed as programmable for new sounds, not only replication.

In other words, as I've written somewhere before: Harmor is the "Minimoog moment" of additive synthesis. Or, in Star Trek analogy, the first warp-capable craft that can be controlled by regular pilots - not only the engineer who built it



I think that an attempt at successful additive synth nowadays must at least match Harmor in capabilities, among which are:

- 516 partials
- animation of pitch and amplitude of every partial
- ability to set partials to inharmonic frequencies (for example 4 partials per semitone)
- freely drawable "filter" shapes
- extensive possibilities for shaping and modulating various parameters (importance of this cannot be overstated)
- re-arrangement of frequency-domain modules



Additionally, I'd wish for, among other things:

- animatable filter contour states, so one could animate through filter curves, comparably to scanning a wavetable but detached from pitch of the note; another point of comparison could be paramertic EQ with ability to "scan" through EQ presets. Harmor can do something like that via crossfading between its two "filters", but the resulting "two frames of animation" are nowhere near enough for complex sounds.

- animatable phases of partials (to my knowledge, Harmor lacks this in synthesis mode)

- polyphonic per-note animation for most parameters including timbre. This is likely very complex to implement and would effectively require per-note instances of the entire synth, with possibility to control all parameters which are not per-note as groups. It is also very hard to realize well from UI perspective.

- good user interface for animation (wavetable timeline in Vital is a step in right direction, I think)

Those would allow to simulate phenomena such as per-note timbres and resonances of acoustic instruments, formants of different vowels , etc.; as well as enable similar possibilities in creative synthesis.



***


I'd be happy to share ideas like that in more detail; that said, most would sum up to "like Harmor, but more animatable and polyphonic"

Personally I lack the expertise to realize the additive synthesizer I dream of, but I'll gladly contribute observations and ideas to any project that dares to take additive synthesis forward.

No offense to Image-Line developers intended. I rely daily on Harmor VSTi and am happy to evangelize it, but I also think that it can - or rather, must! - be surpassed eventually


***

Sorry for the long essay. As evident, additive synthesis is one of my favorite subjects in this lifetime Hopefully this wall of text is of at least some value.

[JSBach1685]

That is an extremely thoughtful reply and you covered everything.
It is obvious that you have plenty of practical experience from the way you replied.
Might as well take another look at Harmour then. Been a while since I've used it.

All I was doing the whole time was nodding my head in agreement while reading haha.
There is really nothing else to add upon what you said apart from yet MORE features.
Thanks a ton for that. You may not think it but that was very helpful.

Feels so good to find another muso that thinks the same way!
This is the complete opposite of reddit.
cheers mate

[JSBach1685]

at least thirty harmonics? that ought to strain the most robust fft implementation

but of course they can't do anything anyway

allegedly


holmes go easy on youself

For analyzing the sounds I have developed a method that involves no FFT. It instead dissects the sound with high-Q notch and comb filters. This makes up for all the shortcomings of FFT at the cost of more processor power. It works fine on a new PC, but it would probably struggle on a smart phone.