Soundtheory GULLFOSS computional auditory perception EQ

[yellowmix]

Very impressed. Low CPU usage, great and easy workflow, excellent results. I gave it some nasty tests to see how it works.

I created sine wave partials octaves apart, then planted a sine wave nearby at half power. I adjusted recover and tame to extremes and I could see that tame scoops the partials and partial bystander with relatively narrow bands, and recover does the same to non-partials in the other direction, and it kind of adheres to Fletcher-Munson (scoops higher frequencies by about 2db more). It's got just a bit over 1024 samples latency so there's obviously an FFT happening and it's precise enough given the EQ bands. Very tight coding! I want to try smaller frequency intervals and see how precise those bands can get. I also want to run it on actual material over the next few days and compare it to stuff like Ozone's Master Assistant (though this does not do limiting so only the EQ aspects).

My only gripe right now is the inability to resize the plugin. It can be tiny on larger screens.

[v1o]

Any DSP gurus here? Is there anyway of testing if Gullfoss actually does what it says it does, how can we be sure this isn't just a FFT based dynamic EQ under the hood along the likes of Melda and TDR.

[martinjuenke]

Any DSP gurus here? Is there anyway of testing if Gullfoss actually does what it says it does, how can we be sure this isn't just a FFT based dynamic EQ under the hood along the likes of Melda and TDR.

Don‘t ask such questions! Just believe in MAGIC!

[plexuss]

Any DSP gurus here? Is there anyway of testing if Gullfoss actually does what it says it does, how can we be sure this isn't just a FFT based dynamic EQ under the hood along the likes of Melda and TDR.

It might be. After all, its a... dynamic EQ. It just needs to have many fast good sounding bands and hooked into a masking detector maybe with some dynamics processing. I think the value here is in the "audio issue" detection, hook into the dynamic EQ and the quality of the EQ. All of which sound like they are tuned and tweeked very well.

[plexuss]

Here's something to try: This helped give me a better sense of what Gullfoss is doing, how it reacts. I mixed a dry simple drum machine loop with a full kick and some toms along with everything else, with a 90Hz sine wave. I mixed the levels so they were balanced. I put Gullfoss on this and it become clearer how Recover and Tame work. Bias didnt have that much effect, but some. Boost appears to be a tilt. The idea was to have one musical track mixed with a tone that would case some masking in key bands and then see how Gullfoss deals with them. Make sure to loudness match!

[lukejwarren]

Really interesting plugin, have got some great results working with the demo. Mostly used quite sparingly, but the clarity and openness it helps me achieve is fantastic

[yellowmix]

Can someone test if bypass truly nulls? I'm finding it doesn't. It's imperceptible to humans but I'm seeing differences at -144db.

[martinjuenke]

Can someone test if bypass truly nulls? I'm finding it doesn't. It's imperceptible to humans but I'm seeing differences at -144db.

Seeing, but hearing as well?

[martinjuenke]

Can someone test if bypass truly nulls? I'm finding it doesn't. It's imperceptible to humans but I'm seeing differences at -144db.

Perhaps it‘s the build in analog noise floor which makes it so pleasant for some...

[yellowmix]

The manual says it's supposed to be a true bypass, though computation is not stopped to keep CPU usage (and I assume the latency) consistent. I ended up comparing renders then comparing the 32-bit float values in a hex editor and there are minute differences. But I want to make sure it's not latency compensation messing with it.

[garfield78]

Seem than Gullfoss make same thing than Hornet 31 with "continuous" mode( https://www.hornetplugins.com/plugins/hornet-thirtyone/ ) .But gullfoss is more expensive .Both are good , i have already buy hornet31 there is some month ago , but i have buy gullfoss for pc too , héhé.

[flocked]

Seem than Gullfoss make same thing than Hornet 31 with "continuous" mode( https://www.hornetplugins.com/plugins/hornet-thirtyone/ ) .But gullfoss is more expensive .Both are good , i have already buy hornet31 there is some month ago , but i have buy gullfoss for pc too , héhé.

Not at all. What Hornet31 does: The “Auto EQ” algorithm compares the average peak level of the frequency bands with each of the bands and computes the necessary gain to apply to the EQ to bring the level of that specific band to match with the one of a carefully designed ideal response.
Gulfoss uses a computational auditory perception model (based on human auditory perception) to understand which audible elements are competing for attention.
Gulfoss uses a much more complex algorithm.

[plexuss]

Something interesting that I found about Gullfoss is that the better audio sounded before processing with it, the less it seemed to do. The more the audio could benefit from further work, the more Gullfoss affected it. This is not usual behaviour of a plugin but makes sense based on the intentions of it.

[meloco_go]

Any DSP gurus here? Is there anyway of testing if Gullfoss actually does what it says it does,

Just use your ears?

how can we be sure this isn't just a FFT based dynamic EQ under the hood along the likes of Melda and TDR.

You need to check how it works on the bass. FFT "bins" are equally separated in frequency, and the latency of an FFT algorithm depends on the size of the bins. So you can't have FFT based EQ to work on a frequncy, say, 50 Hz and have latency less than about 20 ms.

[Ploki]

i'm usually on the technical side of things, but i compared Gullfoss to a few and as it stands now it sounds pretty unique.

Supposedly the closest is Zynaptiq Unfilter