The curse of Spectral

[whyterabbyt]

Ever noticed how cursed spectral plugins seem to be? Generic spectral processing is about the rarest category of effects plugins, yet somehow the ones that are developed never manage to survive.

NI Spektral Delay and Vokator : dropped by developer
Anarchy Effects : only saved by being open sourced
Spectral Monkeyage : dropped by developer
Darrel Barrel DtBlkFx : only saved by being open sourced
Delaydots SoundDesigner and SpectrumWorx : developer ceased operations
LittleEndian SpectrumWorx : developer ceased operations
izotope Spectron : dropped by developer

A bit worrying, they seem to have a much shorter lifespan than other plugins. Im just hope SpecOps bucks the trend....

[Sampleconstruct]

Fortunately GRM have been around for a while and also Melda seems to be going strong.

[Xenakios]

I could think of a few reasons why a spectral processing plugin might ultimately commercially fail even after some initial excitement from users :

1) Spectral processing usually adds a hefty amount of latency/CPU load.
2) Spectral processing often adds particular kinds of artifacts into the sound.
3) Spectral processings might not be that useful to use often in mixes.

All those factors could contribute to the product not selling well in the end. 1) and 2) might even cause additional customer support expenses.

[Ah_Dziz]

Also most of the major products from that list used code licensed from others so when the license expired they dropped the product. I would wager spec ops will stay around as long as Unfiltered Audio does and they don't appear to be going away.

Also if your looking for fun spectral stuff, I'd recommend the soundhack plugins.

[whyterabbyt]

Also if your looking for fun spectral stuff, I'd recommend the soundhack plugins.

Yeah, got them when they came out for PC; Mac development seems to gets priority though.

[CinningBao]

I think one of the biggest problems with spectral DSP is that it's quite difficult to sell something as complex as FFT products to people who don't really know what it is. UA SpecOps _may_ have appeared to buck that trend, but it doesn't appeal to me as much as the modularity of SpectrumWorx or the diversity of the GRM collection. Which surprised me, since I'm a bit of an FFT whore.

I'm also not sure many of the companies who release plugins using Spectral technology actually know how to invite users onto the weirdness of frequency domain magic; I've tried explaining and showing much FFT tech to fellow musicians and technically minded folks, and I'd say 20% of them actually understand what it is, and how it works, and with that, how to use it as a creative tool. It just floats over most peoples' heads.

From looking at the edges of FFT programming, I would also speculate that it takes more time and resources to deliver a solid FFT DSP effect device than it does a 'standard' (for want of a better word) DSP effect. And with that in mind, creating/designing an audio effect, which is future-proof seems quite complex, possibly more so than (looking for the right word) 'regular' non-FFT DSP. For instance, Spectrumworx 1&2 users requested a simple envelope follower as a modulation source.. Aleksey (SW1 developer who was extremely communicative and helpful) couldn't implement any kind of modulation system at all. Nothing. He was the only one working on the code though.. same goes for DtBlxFX and the Anarchy Effect series; small developer count, probably with a fixed idea of the end result during the programming process. Ultimately the end product can't change too much from the initial idea (save for diffierent modules/algorithms) since it would mean modifying the core structure too much, and of course all the dependancies. For a small-man team, this could get nightmare-ish very fast.

Xenakios point 2 is interesting.. I've heard many FFT algorithms, and all the different ways developers try to handle the deconstruction and reconstruction of the windows/blocks, IRCAM have been working on it for years do they've develped beautiful sounding algos.. but all the different ways a developer can choose to code the window edges, handle the phase bins (some devs just ignore them! I think that includes Xenakios since preservation of the phases in the new-fangled PaulStretch would lose the blurring of the windows, on which it thrives), all that 'stuff' changes the perceived quality of the sound. I mean, I LOVE GRM tools which offer block sizes of 2/4&8 or the super large 16,384.. that's part of the magic of FFT for me.. I suspect though, most users aren't even aware of this as an option.

So yes, FFT tools, difficult for developers to commit to, difficult to sell to consumers, difficult for consumers to understand since they are so technical in the first place.

Just my two-penneth

--edit missing word

[cron]

I found the marketing surrounding SpecOps to be quite revealing here. Unfiltered Audio were really focused on the FSU/extreme end of things when posting sounds they liked from beta testers, but when the plug actually released, Plugin Alliance put surprisingly strong emphasis on using it as an enhancer or mixing tool. Around half of the Soundcloud sound examples demonstrated the more polite end of its range, compression/enhancement and the like, while the launch video was mostly reaction shots from producers with a banging pop track in the background. There are only around 2 or 3 actual sound examples in the video. Marketing this fundamentally experimental plug-in to a mainstream audience.

I'm all for it if that's what it takes to sell plug-ins and keep the niche plugs I love coming, and SpecOps can be used as a subtle sweetener of course (though I doubt it'll be many people's first port of call for that particular task.) I guess that courting the mainstream is a necessary reality of plug-in development these days, and the niche FSU crowd just isn't big enough to keep the money rolling in thus sustaining continued development.

[Xenakios]

handle the phase bins (some devs just ignore them! I think that includes Xenakios since preservation of the phases in the new-fangled PaulStretch would lose the blurring of the windows, on which it thrives)

Yeah, no plans here to start changing the algorithm to do anything "correct" with the phases. How the PaulStretch algorithm just substitutes random numbers for the phases is an integral part of the algorithm. Admittedly that also causes the "gives every sound the same kind of artifacty flavor" issue...But it's what it is. The resulting sounds may or may not be useful/interesting for someone. The processing is definitely in the FSU category, not suitable at all for light sweetening/enhancement/time or pitch adjustment.

[ChamomileShark]

I got the SpectrumWorx demo just before they went - bit of a shame because I would have bought it. I had Specktral Delay and Vokator until Cubase went 64bit only (but I just got Bitwig 8 track as part of Roli package - they may work in that?).

I now have the GRM creative tools (with an eye to adding to those when I can afford) and the SpecOps.

But yes, they do seem to be a bit rare - and often a bit more pricey.

[VariKusBrainZ]

Let's not forget CDP, little hint for Xenakios

The Renoise integration is quite nice

[Xenakios]

Let's not forget CDP, little hint for Xenakios

They also had to just let it eventually go free and open source...I doubt they ever had significant sales going on. Them going open source also hasn't caused a major impact. The code is very difficult to change and adapt to work in any other way than it has worked so far. For example their spectral processings depend on precalculated analysis files and only support mono processing. Those things are not at all trivial to change/fix.

(And I haven't forgotten about the CDP front end stuff, just taking a proper break from thinking about it. The λ software also aims to be more than a CDP front end, so there are all kinds of complications that need to be rethought carefully.)

[plexuss]

Specrtal processing tends to impart a particular sound to the audio, no matter what the tool. Be it a synth, a filter, reverb, etc - they all have this... sound. The sound is seductively interesting but not very musicial when you get down to working with it. It can be effective in certain circumstances but in general spectrally processed audio tools are not very versatile. Not to say the spectral sound is not good, or anything like that, just that they in general have a sound signature that is unique and overt which can limit their use. But it depends on the artist....

[Russell Grand]

I'll add Linplug Spectral to the list. Granted, technically Spectral died due to Linplug closing up shop, but the fact remains their Spectral synth is dead.

[aMUSEd]

Not really surprising

[Neon Breath]

But you can easily make a list too of the devs still there with their spectral plugins...

GRM Tools
Michael Norris Soundmagic Spectral
SoundHack
Melda Production
Expert Sleepers
Artificial Audio
Unfiltered Audio