The curse of Spectral

[_al_]

reafir comes very close, but i dont use that company for personal reasons, so the search goes on

Anti-Christian?

You know, it's funny. I'm not even Christian, but at the same time, there's some very dark stuff happening in this world under the surface, so I try my best to not offend any God who may exist

So when I see a plugin dev called Jesusonic, I just see an unnecessary level of public disrespect that I want no part of.

Yep, I'm weird.

How did you use sonic foundry noise reduction for sound design? If you change fft resolution to 128-512 and move any sliders it just gives noise or metallic sound...
What's the purpose to split sound into transients and non-transients?
Does anyone have soundhack vst plugins for linux? (Not for pd)

Yeah, the low sizes sound horrible.
But 16k was the magic setting for me.
When you set it to play noise only, you would get the most eerie and beautiful sounds from it... so when you morph between the noise-only and the dry... it's like this ethereal ghostly reverb that was unlike anything else on the market at the time.

But hey, this is 2025. I've even got some similar effects from the likes of Shaperbox 3's new fft fx.

Still though... I don't have anything new that gives the ease of use AND accuracy that SF's NR had.
Funny to think that I might of been one in just a handful of people who were using the plugin like that? I heard nothing in any electronic music back then that had a similar sound. (this was like 20 years ago or something)

I know this is super old but if you are having trouble replicating fft based noise reduction you may want to look more closely at how it's achieved. It's basically a spectral gate/expander and also called spectral thresholding. It's attenuating frequencies that fall below a certain threshold level in the same way a broadband gate does. The main settings that will change the sound of spectral gating (and most spectral processing) are the fft size (directly effects the number of frequency bins) the overlap (in order to deconstruct and reconstruct the signal properly there are usually multiple ffts happening in parallel and offset in time while being multiplied by a windowing function) and then the way phase is handled. The difference between a 256 bin fft with a hanging window and overlap of 4 and a 2048 bin fft with a rectangular window and overlap of 1, with all other settings set up the same, is massive.

ReaFIR, and melda MSpectralDynamics both have reasonable amounts of control over all this stuff. You could also easily build your own in something like pure Data. There is a tutorial that will have you build a spectral thresholding processor.

I always wondered if Spectral dynamics would give me the same level of control. Pretty sure I did try the Lite version, and was pretty disappointed though. Didn't come close to SFNR or Reafir.
If you think the full version might get there though, let me know.
The beauty in SFNR was that you could carefully craft every little detail of the sound using it's graph (basically like using an eq).

I do have many of those plugins you listed (used DTBLKFX quite a lot, but the original version that didn't have the weird pop up sliders)

Currently, my fave is MTransformer. That one is VERY similar, but I haven't used sfnr for so long, I can't remember how close it is.

But honestly though, when you start getting into the maths of it all, I start to lose interest fast. Maybe just too deep for me.

And still though... spectral fx effects are my favourite.

[Dirtgrain]

The minute they stop proselytizing, going on missions, pushing their religion into government, and knocking on our doors on Sunday or whenever--showing all that disrespect--that's when you should worry about their feelings. Not that you should go out of your way to disrespect them (don't provoke Charlie Hebdo kind of religious furor), but if someone else does, why say anything about it at all?

[Ah_Dziz]

I'm not familiar with the exact noise reduction you're referring to. Soundhack has a drawable threshold line and can capture that threshold, MSpectralDynamics does too. Both can do all types of strange things like turning up frequencies when they drop below the threshold which brings crazy background sounds up and with adjustable attack and decay settings you can make all types of crazy and beautiful sounds from almost any input. I've been a big fan of playing with spectral junk since I first played with cool edit 96. I have bought almost every spectral effect to hit the market. The GRM tools stuff is extra cool too. Spectral warp and evolve (I think) are two of my favorites. Sometimes something simpler like SpecOps is what you want/ need.

[jacksmyphs]

Which version of sfnr did you use? I installed one from 1997 and it produces periodic clicks with period depending on window size. It also seems to fail to output only noise..
I remember testing spectralgate and it produced noisefree sound like you could extract ambience from recordings which was free of any artifacts. So seems a good replacement for clicky sfnr.

[jacksmyphs]

I guess you mean freqwarp and evolution from grm. I also thought about them as a replacement to sfnr. But they only extract a wave from incoming audio. How do you further process that wave? Some resynthesizers like magenta?

[Ah_Dziz]

I've never used sound forge for noise reduction. I used cool edit for file editing and that's where I was first introduced to fft processing audio. I don't have any idea what magical things Sonic Foundry did in it's noise reduction algorithm. I also have no idea what you mean by "they only extract a wave from incoming audio". One periodically interpolates between spectra, while allowing for interesting blending between noise and non noise components of a sound and for scaling the pitch and frequency of those spectra and the other is basically a way to nonlinearly remap frequencies the way a wave shaper nonlinearly remaps amplitude.

[ChamomileShark]

seeing as this thread has come up again, I'd add the free Spectral Suite to the list.

https://www.andrewreeman.com/spectralsuite/

I think that "Morph" is the same process as GR Tools "Warp" in Spectral Transform bundle. Anyone can confirm?

[martiu]

if you want good fft noise reduction plugins (free on top of that)
you can try - voiceofsnow https://www.kvraudio.com/product/voiceo ... ndevstudio
ASAP Spectral Clipping (you need reg to the forum to download the bundle)- https://forum.ircam.fr/projects/detail/asap/
and Spectral Gate from andrewreeman spectral bundle - https://www.andrewreeman.com/spectralsuite/

[martiu]

i personally enjoy spectral plugins for manipulating voice or creating interesting sounding noise or drones

[jacksmyphs]

I use arboretum ionizer for eqing and noise reduction. So i guess it can be used instead of spectralgate by soundhack. Perhaps i misunderstand smthbut i understand it the following way (in stages):
1. You extract waveform from incoming audio by an fft tool like grm warp, evolution, sfnr or any other fft plug which converts to fft and allows regularization i.e. extracting mostly harmonical content by setting number of bands and/or other way. Usually harmor or wna additive and alike extract harmonics when you load image or wav.
2. So having obtained a waveform you need to feed it into a resynth for midi controlling pitch, poliphony and other usual processing.
3. Feeding stream from resynth into usual fx plugins for additional sweetening the sound.
The idea behind stage 1 is getting the initial waveform with full control via sfnr or similar plugin. Of course you can directly feed soundstream into a resynth like magenta or ribs but this way you lose vast manipulation possibilities. Thanks for the links to nr soft. I will test it vs ionizer.
The described steps work as a modular concept where you subsitute soft for different outcome.

Harmor would be much better if it could accept live stream like ribs or magenta. The described concept solves some sought after aspect of sound which constant random modulation and detuning as it happens due to initial changes in initial audio stream itself. This way you get results like from famous absynth or wna additive. You get lsd psychedelic sound which still preserves some characteristics of initial audio and can be layered on its top. Was your idea about sfnr similar to this? It does produce absynth like sound which is cold, rigid and reminiscent of 90s sound. I suspect that replacement of sfnr by any later plugin will kill sound easily. If you use arboretum hyperprism pack you can get an idea about sound of 90s. Plugins from 00s to present day give much softer result as if adding some antialiasing. No wonder why you can't find replacement to sfnr...

[jacksmyphs]

Btw i avoid rea plugs and mixing (summing) in reaper for the aforementioned reason i.e. softening of sound and i sum in some old daw with reaper only exportings tracks as stems. If mixing in reaper the result is softening and loss of highs. You are not alone in taking precautions against those products. If its placebo then our placebos coincide...
Just a thought: software since 00s was started to be in a way calibrated for brownian noise like harmor. When using wna additive or some other soft from pre2005 era you often come across overloud painful high frequencies but this never happens in post 2005 software. It's very difficult to get painful highs with it unlike with ionizer or old grm soft.

[jacksmyphs]

Extract = deffine = set (for initial waveform). I used extract wording due to nature of resynths to extract harmonic content.

[_al_]

The minute they stop proselytizing, going on missions, pushing their religion into government, and knocking on our doors on Sunday or whenever--showing all that disrespect--that's when you should worry about their feelings. Not that you should go out of your way to disrespect them (don't provoke Charlie Hebdo kind of religious furor), but if someone else does, why say anything about it at all?

Actually, I think you are blaming God for the acts of men.
All those things you mention are done by people. When they burned women alive as witches because they dared to show a some individuality, do you really think there was a God smiling down saying "oh well done, that showed her"?
A real man of God will show nothing but love, compassion and respect.
He will respect the fact that God gave us free will for a reason.
Though not Christian, I HAVE read the bible, and it's made clear, he gave us free will because he wants us to choose him over the world.
Anyone who forces anything upon you in the name of religion, is not doing it in the name of God, I can assure you.
And honestly, most people who call themselves Christian don't even come close imo.
I think there are very few true Christians left in the western world now.

And if someone knocks on you door on Sunday morning, yep that can be irritating as f**k, but you only have to smile, shake your head and close the door. That person will then go on spending their whole day on something you consider pointless and futile, but you on the other hand will already be back in your seat, enjoying whatever it is you do on a Sunday morning.
Was that little knock on the door REALLY such a big deal, considering that they are still out there getting doors closed in their face, because they believe they are trying to help people?

[_al_]

This way you get results like from famous absynth or wna additive. You get lsd psychedelic sound which still preserves some characteristics of initial audio and can be layered on its top. Was your idea about sfnr similar to this? It does produce absynth like sound which is cold, rigid and reminiscent of 90s sound. I suspect that replacement of sfnr by any later plugin will kill sound easily. If you use arboretum hyperprism pack you can get an idea about sound of 90s.

Hmm, it kinda sounds like you know what I mean, but actually, I didn't consider it as a 90s sound at all, because there was literally nothing out there that could do this to audio.

To try and explain in more detail...

Imagine pouring hot lava over a sound, so that it starts to melt away the most delicate areas of the sound.
The more lava you pour, the more it melts away, eventually leaving just a skeleton, which would be the opposite of the loudest fundamental frequencies.
but because the fft size was set to max (16k), it wouldn't be little bits of tonal noise at certain frequencies. It would instead be like a slowly morphing blob of goo, with bumps and lines that appear and disappear over the course of the sample audio.

Heard on it's own, it sounds very creepy and ethereal. Like a dissonant squeaky gate kinda sound. But when you start to mix in the original signal, you start to hear these little ghostly squeaks slowly fill out, as it starts to become the original sound again.
There will be a 'magic' spot in the wet/dry, where it just sounds completely fking weird and unnatural. And back in the 90s, there was really NOTHING else on the market that gave me this type of sound.

But sfnr could take it further than that, as it had the EQ graph. (Sonic Foundry Noise Reduction, if anyone's wondering)
You could use the graph to seek out and boost specific areas of the sample, leaving a result that actually didn't even sound like it came from the same sample.
And yet, when you slowly mix the dry back in, it still morphs perfectly back, because it was indeed frequencies that were present in the original.

And I still don't have anything that does this comfortably.
MTransformer does get really close, but it lacks the detail that I was getting from sfnr, because that had the "Capture Noiseprint" button.
Once you've clicked Capture Noiseprint, you could then go on to use the EQ graph to fine tune that noise print. Like you are drawing in the sound, if you like.

Was always such a fascinating approach to "out there" sound design, and I was always annoyed and quite shocked that it was never developed further by any other company.
But I guess that was because I wasn't even using the plugin for what it's meant for.
I was doing the opposite. Keeping the noise and shaping it further.

[jacksmyphs]

Really interesting info. It seems i have used it the wrong way as i have not pressed the capture noise button. Probably it's the cause why i failed to output only noise. I did not see any eq either except some high cut or shelf. Indeed it often happens that you get stunning results by using inappropriate approaches. There is element of fortune which opens the gate to another dimension.