Overestimated synths?

[ghettosynth]

Sure there's resynthesis, but thats basically sample mangling and rather missing the point.

I strongly disagree with this. You can't just casually dismiss one of the features that best shows off the power of additive synthesis just because doing so is convenient for your argument.

Additive resynthesis, and Harmor's implementation of it, is special precisely because it isn't sample mangling. Once the sample is resynthesized, it's part of the synth engine. It's no longer a sample, but a collection of parameters. Suddenly it's manipulatable at a fundamental, sound-generating level, with as much control as if you'd made it from scratch, as much as you'd have with any subtractive or FM equivalent.

This also helps reduce the issue of the extreme complexity of additive that ghettosynth mentions. If the sound you want to make is close to something that already exists, you can get most of the way there without having to reverse engineer the starting point yourself.

Yes, but, there is validity to the point that subtractive processes add a necessary dimension to additive as well. While resynthesis can give us a parameterized model of a sound across spectral dimensions, it cannot necessarily give us a parameterized model in terms of the dimensions that may interest us, e.g., the particular dynamic harmonic adaptation that a particular filter provides. That is, we can manipulate the model, but we may not be able to manipulate it in the same way that we can when we combine it with subtractive processes.

[AnX]

What are some of the most overestimated synths in the world?

Do you mean overrated ?

[ghettosynth]

What are some of the most overestimated synths in the world?

Do you mean overrated ?

Well, if so, that would be a different, and much less interesting IMO, thread. To me overrated is more of a post-perception where the synth becomes popular and then people think more highly of it than they should. Overestimated, on the other hand, is a post-analysis of a pre-perception where either the manufacturer or the public had more faith in a product than was deserved at the time of release.

Volkswagen "The Thing" = overestimated
Volkswagen "Bug" = overrated

[nineofkings]

Yes, but, there is validity to the point that subtractive processes add a necessary dimension to additive as well. While resynthesis can give us a parameterized model of a sound across spectral dimensions, it cannot necessarily give us a parameterized model in terms of the dimensions that may interest us, e.g., the particular dynamic harmonic adaptation that a particular filter provides. That is, we can manipulate the model, but we may not be able to manipulate it in the same way that we can when we combine it with subtractive processes.

This is a bit of a circular argument though, isn't it? If the specific sound you're trying to create is the defining characteristic of subtractive synthesis, sure subtractive synthesis is gonna be the best way to do it. If you want to create guitar sounds, your best bet is to use a guitar. This doesn't necessarily mean one method is superior to the other or even a necessary component, just that it is distinctive.

[ghettosynth]

Yes, but, there is validity to the point that subtractive processes add a necessary dimension to additive as well. While resynthesis can give us a parameterized model of a sound across spectral dimensions, it cannot necessarily give us a parameterized model in terms of the dimensions that may interest us, e.g., the particular dynamic harmonic adaptation that a particular filter provides. That is, we can manipulate the model, but we may not be able to manipulate it in the same way that we can when we combine it with subtractive processes.

This is a bit of a circular argument though, isn't it? If the specific sound you're trying to create is the defining characteristic of subtractive synthesis, sure subtractive synthesis is gonna be the best way to do it. If you want to create guitar sounds, your best bet is to use a guitar. This doesn't necessarily mean one method is superior to the other or even a necessary component, just that it is distinctive.

Not at all. First, define subtractive synthesis as you mean here? Traditionally it means starting with largely fixed harmonic rich sources and then using filters to remove the harmonics that you don't want. I'm not talking about that at all so the examples that I'm imagining cannot possibly be related to "the defining characteristic of subtractive synthesis."

You suggested that re-synthesis can be a good place to start to obtain a sound that is similar to a sample, and I agree. Now, once you have resynthesized it then the notion of "similar" necessarily implies that you want something slightly different. What that difference is will define what is the best approach to get there. If what you want is something that is slightly different in terms of the spectral parameters then maybe you can use a pure additive approach and make some adjustments. If what you want is just a filtered version of the original sample, then you've wasted your time with resynthesis because a sampler with a filter would have provided the optimal path. But, what if you want a sound that is slightly similar with respect to the additive parameters, but also has the sonic footprint of a particular per/voice filtering process? Now your sampler is not ideal, neither is additive alone as I don't believe that you will achieve that ideal filtering, and certainly you are very far away with a pure subtractive process. So, here, the ideal process is hybrid, re-synthesize, adjust, and filter.

Suppose your goal is to manipulate a guitar sample using resynthesis and then apply the sonic footprint of a TS-808 per string (voice). This is a perfectly valid synthesis thought process. Please tell me how you would obtain the specific dynamic character of the TS-808 and adapt that to an additive process? Developers and pedal developers alike have spilled plenty of blood sweat and tears over the tiniest of details of the sound of that pedal. I don't think that you can dismiss the validity of that work and say "I can do that with a purist approach." I contend that it can't be done in any practical sense at all.

Sampling with the TS-808 will not capture the dynamics of the pedal at all, so that's out. You could approximate it with an additive process, however, if approximations were good enough there wouldn't be so much interest in models of specific filters (used here in a broad sense).

One can't talk about FM as driving a different thought process with respect to synthesis and then dismiss that thought process for other methods and ideas. If I can break a sound down into its separate parts to reassemble them with an FM synth then I can apply that same thought process to synthesis in general and choose the best model for each component of the sound across all of the dimensions that interest me.

[nineofkings]

I feel like you have something in mind when you say "a particular per voice filtering process" where you lose me. To me, this means "the characteristic sound of a filter," which is what I meant when I said the defining characteristic of subturactive synthesis. If that's wrong, can you give an example of what you mean?

EDIT: original post was edited and answered this question. Two different uses of the word "filter."

[aumordia]

Resynthesis is a variation of sampling -- you're starting with a sample of something else. I always thought that was one of the nifty things about harmor's additive engine -- that it could do sampling. I'm not sure how resynthesis exists without sampling. Could you show me how to resynthesize without a sample?

Really it can be quite fun to subject a sample to Harmor's additive engine -- but I'll tell you what isn't fun: trying to additively synthesize something realistic and complex, like a bowed string. This is how additive synthesis is overestimated -- in theory, it can do this. And resynthesis proves that it can do it in practice as well. However, it is impractical to program such a patch, hence people resort to sampling other sounds.

Harmor actually lets you do this all over the place -- you can create an envelope out of a sample -- but this is more proof of the power of sampling than of additive playback. I personally don't like working with samples -- I am attracted to pure synthesis -- but sampling as a paradigm really has delivered on its promises. Harmor was made much more powerful over Harmless by the inclusion of sampling, which is telling.

[nineofkings]

It is not sampling though. It's reconstructing a sample using synthesis. In this respect it's identical to placing all the sine waves by hand; it just analyzes and does the work for you. That's why it is an inherently different process, even though it uses samples.

Sampling gives you a recording of a sound and that's it. You can't easily change the spectral characteristics per voice, or the pitch and rate independently, or the formant signature per voice, because you're working on a recorded sound. With Harmor, you're working not on recorded sound but on a synthesized reconstruction, meaning you can change all of those things per voice directly, rather than working on an actual fixed recording.

[ghettosynth]

I feel like you have something in mind when you say "a particular per voice filtering process" where you lose me. To me, this means "the characteristic sound of a filter," which is what I meant when I said the defining characteristic of subtractive synthesis. If that's wrong, can you give an example of what you mean?

Well, I gave an example, but I'm not sure whether this matters. If what you want is the characteristic sound of a filter applied to an additive or FM sound, then that's not a pure subtractive method, nor is it pure additive or FM, and none of those methods alone will achieve your goals in practice.

FM8 really does deliver on this. Browsing through the patches demonstrates that filters are useful in FM without necessarily needing to simply "impart the sound of the filter" or to simply "filter the output" in a trivial way.

I see no advantage beyond that which is enforced by technology to staying in one synthesis method when a hybrid approach is more appropriate.

The point that I was making with respect to additive resynthesis is that the parameters that are easily derived may not match the parameters that you need to achieve your synthesis goals. Consequently, you cannot use the fact that you are able to derive those parameters as a defense for pure additive synthesis over a hybrid model. This is no different than the FM argument. You have a set of parameters and a goal to achieve. If you can't find a mapping between your given parameters and your goal parameters, you cannot achieve your goal with your chosen method. That doesn't make it circular reasoning because I gave examples that are dominated by subtractive tools. Those ideas are very valid and form a huge part of what we do in music production.

[nineofkings]

Yes - I saw your post before the edit with the example given. Confusion over the word "filter." I was expecting a synth-head like yourself to be using it in a synth sense rather than a generic sense

[aumordia]

So when you drag and drop a sample onto Harmor, and then it plays that back, that's not sampling, because the underlying technology is additive?

Hmmm... I wonder if you're still legally obligated to get your samples cleared if you use this technique.

[AnX]

What are some of the most overestimated synths in the world?

Do you mean overrated ?

Well, if so, that would be a different, and much less interesting IMO, thread. To me overrated is more of a post-perception where the synth becomes popular and then people think more highly of it than they should. Overestimated, on the other hand, is a post-analysis of a pre-perception where either the manufacturer or the public had more faith in a product than was deserved at the time of release.

Volkswagen "The Thing" = overestimated
Volkswagen "Bug" = overrated

Ah ok, so two companies complete ranges spring to mind here...

[ghettosynth]

Yes - I saw your post before the edit with the example given. Confusion over the word "filter." I was expecting a synth-head like yourself to be using it in a synth sense rather than a generic sense

Ah sorry, no, I may come across as a synth-head, but, my thoughts tend more towards the abstract. There's an unfortunate clash of ideas here since, strictly speaking in terms of a signal processing concept, a filter removes components. However, practically speaking, filters also add components. Sometimes the sonic footprint of a filter is really more about particular kinds of emphasis than removing components, per se. FM of a great resonant filter with whatever source you are "filtering" often sounds great with the filter wide open. You're not going to easily achieve this sound with FM or additive processes.

A TS808 isn't strictly speaking a filter, but it adds components and remove components in a way that may or may not map to additive parameters.

[ghettosynth]

So when you drag and drop a sample onto Harmor, and then it plays that back, that's not sampling, because the underlying technology is additive?

Hmmm... I wonder if you're still legally obligated to get your samples cleared if you use this technique.

There is a famous case about this regarding resynthesis of some works by the Beatles. It was ruled that resynthesis is just an elaborate copying method and, consequently, clearance is still necessary.

[aumordia]

I learned something new today.