This, with pink noise input.


Looks just like two-band EQ...
Or pink noise curve applied to a pink noise source?
(It should read 6.25 kHz, my mistake.)

 

Last edited by a moderator: Oct 25, 2020

But now it's three, with just TAME included:

Or, is it?

 

theres a little audio engineer inside gullfoss, asking too many questions will annoy him and cause Gullfoss to stop working as well

 

It should be three, at last:

 

But then, what's this (with the same input file:

 

@pratyahara one is doing spectral balance and the other resonance suppression

 

@Kwissbeats Exactly!
People should stop willing to compare everything multiband, Gullfoss & DSEQ2 are two different processors, so you won't get the same result obviously...

In short, Gullfoss is an "intelligent" (i-e source dependant) eq aiming to maximise clarity, and DSEQ is an intelligent de-harsher (or resonance suppressor).

As for your question regarding Gullfoss, it's an eq capable of changing its frequency response more than 300 times per second (so a static picture can't be helpful to represent its action).
Gullfoss is adapting the incoming signal with an "ideal" predefine spectral balance based on a model of human auditory perception, adjusting it's action in a very fast dynamic way (and depending on your parameter settings).

Hope this was helpful

 

All dynamic processors are source dependent.
The graphs show that Gullfoss does not 'maximize clarity', but is basically a 3 band EQ with interdependent settings. It does not require any intelligence to compare the input signal's fundamental frequency with some preset curve.
The picture is helpful to represent its action - test pink noise does not change spectrally or dynamically, so it was the same identical curve for 15 minutes, so each screenshot would be exactly the same.
Maybe (*if you trust the developer*) Gullfoss compares the input 300 times a second to a preset curve (what you see in the graphs is really and solely Gullfoss untouched preset curve (adapted to user's choice concerning intensity) So Gullfoss just does adapt the input signal using the constant preset curve 300 time a second (=300 comparisons to the static curve a second), and does 300 static filterings (or compressions, like Voxengo TEOTE does?). Gullfoss isn't changing its frequency response but does the frequency shaping of the input signal.
Nothing in the shown frequency response of Gullfoss shows that it is based on a model of human auditory perception (and which model should it be, does it have a name?)


It is certainly not the Mel scale (see above).
How can user contribute to an "ideal" predefined spectral balance?
On graphs shown in my post it is clear that Gullfoss provides user controls that just make nonsense of that ideal" predefined curve, and only can ruin it.
Developer's stories about 'quantum physics', 'intelligence' (he does not say "AI" - strange, remember? - 'waterfalls' (he does not say he took the frequency fingerprint of it and that he applied it; and he does not say what 'psychoacoustic curve'...
Thank you for your help, but I have an impression you just paraphrased the developer's story.
If you take the curves seriously and analyse them you might get some revealing conclusions, like this:
https://www.kvraudio.com/forum/viewtopic.php?p=7470030#p7470030

 

Last edited by a moderator: Oct 25, 2020

Not quite, I'm afraid. In 'tame' mode only, Gullfoss is supposed to do the same as DSEQ - but it doesn't.

 

Correct. You can see this by boosting a frequency gullfoss is taming. Not much changes in gullfoss until you set up some extreme settings. In tame mode it basically everytime tames frequencies <200Hz. That is why I do not like the tame mode very much. And in general I limit the frequeny response to 200 - 6000Hz because it acts very "static" in higher or lower ranges.

 

I understand your argument, but still doesn't think your starting point is relevant : judging a processor which purpose is to react and change extremely fast to an incoming signal, by feeding it with a static one, seems kind of non-logical to me.
If you feed any compressor with a static signal, then you could assume that there's no difference between fixed value release and program dependant release... is that that program dependant release is a scam?

And yes, dynamic processors are source dependant, but equalizers are not : gulffoss is both dynamic and spectral, that's why there is a point in mentioning that both its dynamic and spectral action are source dependant (I don't consider that being "artificial intelligence" either, but that's part of the Soundtheory's com i guess)

What seems a valid question is the predefine curve the developper has set in his processor: for that you can also use and trust your ears, I think that's the main reason for allowing the user to adjust the recover and tame parameters. It's not a magic box that will make all the right decisions for you.

 

Smoke and mirrors.

 

First, in this way I caught their fundamental EQ curve. And it happens not to be anything esoteric, but just a standard pink noise curve (like that found in Sonible Smart EQ2, for instance). I could not find any 'psychoacoustic' modification in that curve (except making obligatory 'bump' at 16 kHz /to make sound more appealing to customers, I guess?/.

Second, all other features besides basic and logical /having in mind Gullfoss purpose/ 'recover' and 'tame': - ('brighten', 'bias' and 'boost') - just juggle with this curve, independently of input signal, at sheer user's will. If Gullfoss reacts automatically and intelligently according to adaptive and 'psycho acoustically ideal' curve, those additional features mean ruining its job, whatever else?

Third, it reacts to an incoming signal, no matter static or dynamic (if we consider noise as 'static', at all, it is not a sine or square wave) - but solely by blindly adapting it to the pink noise curve.

Fourth, Gullfoss filters, although there is nothing to filter in pink noise, so it works, and does its job; meaning it was not on 'stand by' during measurements.

Fifth, "tame' function works almost symbolically, with some sort of very flattened pink noise curve + 'bump', with mid-band excluded and bass band reversed. It does 'tame' very little in that way in fact. So the main and dominant function of Gullfoss is 'recover,' which in fact makes Gullfoss works predominantly like any other 'spectral enhancer EQ'.

The developer launched one too many fables in order to mystify his product, so why should we trust him? He can't deny the graphs. And it seems, regretably, it is easier for people to take the developer's propaganda for granted than to investigate, test and analyze. Even if somebody did it instead of them.

 

IMO it's like you're comparing 2 different compressors. They both have their own way of working with the signal. Whatever Gullfoss is doing it sounds subtle and great. It cleans up the sound in a pleasant way. Maybe what the graph is displaying isn't 100% true to what the audio is doing? If the marketing around Gullfoss is true there are decades of research into human hearing behind it. Not just what pink noise does to a signal.

Use DSEQ, Gullfoss, Sonible Balancer, Dynamic Spectral Mapper, MspectralsDynamics, iZotope Neutron Sculptor, Soothe, etc, etc however you feel works in your workflow.

Don't reach for a screwdriver when you need to hammer a nail.

 

Well. I did not compare DSEQ and Gullfoss, I just showed the difference between them in one instance, when considering Gullfoss 'tame' (dis)function.
'Pleasant' is subjective, and deserves individual respect, but what is Gullfoss objectively doing to the sound is something else and a matter of rationality.
Results show that technically Gullfoss does nothing special that other dynamic spectral EQs don't do.
The graphs posted are those of delta files (the resultant files after superposition of input and output waves), so they represent the difference that Gullfoss imposed on input file - i.e. the graph is displaying 100% of what Gullfoss is doing to the the audio.
And the graphs posted are just screenshots of Gullfoss's own graphs.

 

... stays in Gullfoss...
oh wait....
Wrong thread, wrong topic... wrong everything

 

&...Wrong usage.

If you know how and what some program does, you'll use use it more properly, have a greater control, avoid mistakes, use all of its capabilities. Otherwise, everything might get wrong.

 

Interesting. Here with pink noise the EQ is mainly moving around the 0dB line, with a slight tendency to apply what's captured in the pic:

 

Yeah. I also do not believe that it moves to a certain "pattern" per se. As I said. You can apply some equalization before and gullfoss is doing pretty much the same as without pre eqing.

I think it simply "tames" the most insignificant and "recovers" the most significant elements of the signal.

And the method is basically stated at their front page: "...equalizer that listens to a signal and decides how to prepare the audio so that your brain can get the most information out of it. The realtime analysis of Gullfoss uses Soundtheory’s computational auditory perception model to understand which audible elements are competing for your attention."

 

I will keep watching this thread to see if you understand it. For me, it's one of those things that I don't try to understand I just move the knobs till it sounds good. Some of those tools are nice too. It's like a slider in photoshop IMO move it one way it might look better. move it the other way it might look better.