ZYNAPTIQ Intensity

[bbtr]

It can be listened to, running audio through it, but half of the forum has tinnitus apparently.

Bah, listening is so passé...


We got Tools Measurement tools.

[zynaptiq]

Just took a peek at the manual: for linear analysis Plugin Doctor uses Delta (first sample 1, all other 0) and Random (noise). For THD uses a sine wave, for IMD - two waves.

That sounds pretty much what I had though would be the test signals that would give results like the ones posted, thank you – I was going to look into the Doc plug myself but haven't gotten around to it yet. Those are valid approaches for the things it is designed to test – not for INTENSITY, or specifically the main algorithm of INTENSITY. If using ONLY the SATURATE section of INT, then the THD analysis would probably give valid results for that section (and yes, the saturate has THD of a very nice flava, and potentially looooots of it, too ).

Here's a (rather accurate) analogy – if you present a facial recognition system with an image of randomly positioned dots and line segments – it will still see all sorts of faces in there. Because the algorithm assumes that there IS a face to be recognized. Or at least it will TRY to find a face within the data. Similar to how the human brain will see shapes in random structures like clouds or ink splats.

Similarly, INTENSITY – which uses technologies also used in some types of facial recognition – assumes that there ARE broad strokes as well as details in the input. For a sinewave, a single sample, and noise that isn't a valid assumption, so behavior may be unexpected if feeding it that kinda stuff.

At the end of the day - *listen* to it. Play with it. Mix into it. Start your project using it. Try different values of the controls, especially: don't necessarily CRANK it all the way up. Sometimes 5% can be a very strong effect. It all depends on your signal. And try to not judge it based on comparing it to a compressor, eq, exciter, or anything like that. It is none of those. There are conceptual overlaps in terms of what you want to achieve with them, but INTENSITY is called INTENSITY and not Z-COMP or UNDYNAMIC for a reason

[mcbpete]

Would you say that the best results of the plugin are when Intensity is applied to:

a) Individual tracks
b) Sub-groups (e.g. a grouped collection of drum tracks)
c) Final 2-bus/mixbus

Also would this be applied before the final mastering or as part of the final mastering?

Cheers

[generaldiomedes]

It seems that I have to create a really bad sounding track first .. and then it will work ... ?

[sircuit]

Would you say that the best results of the plugin are when Intensity is applied to:

a) Individual tracks
b) Sub-groups (e.g. a grouped collection of drum tracks)
c) Final 2-bus/mixbus

Also would this be applied before the final mastering or as part of the final mastering?

Cheers

From my experience until now, the first thing is that the plugin is very unpredictible. I mean, you turn it on, start dialing intensity (until you hear "clarity", "details", something) but you have no idea if before reaching what you expect to hear: it will mush transients or bring unwanted reverb tails (harshness) or distort the kick or bring out other artefacts.

It can be quite predictible in the examples provided and similar: music with no long reverbs, no background sounds, minimal complexity.

On mastered dance music it does nothing or is unnoticed. On unmastered dance music (top40 style) it does something but not enough for a master - so it does less than a master chain but after a proper master chain does nothing...

On individual tracks and sub groups it does something as long as the material is very dry and relatively all the elements at the same level. But the way I have it: the drums submix in a techno track with 3 closed hihats layers and a shaker that is almost unaudible, when I use the plugin it immediately raises my shakers by a lot in context of my mix and i really dont want that.

So I don't know - at the production stage where anything is under control (if I want sparkle I can dial it in, if I want more density I parallel compress or saturate the signal). Tried it even with guitars - an eq boost does the same and it leaves the transients untouched.

I really don't care about the face recognition AI mumble jumble, I care about the sound and for the moment I can't find use for it. I mean anything that sounds pleasant I can do with my tools - it would be nice for zynaptiq instead of writing chapters, just provide proper audio demos: pop, rock, top40, house, disco and show what's better in his opinion. Otherwise we're spinning in circles: we say one thing, the dev says "you're holding it wrong".

[mcbpete]

That's a great detailed response - Thank you

[noiseboyuk]

It seems like a pretty handy plugin to me in that it's a great time-and-effort-saving way to make stuff sound better (but not always, depending exactly on what the stuff is). Where I'm much more dubious is that the marketing around this leads you to believe that something amazing and revolutionary is going on when it really isn't, and the occasional suggestion that you can make stuff sound louder without affecting the LUFS which is I think completely wrong.

To that end, I'd definitely be in favour of real volume compensation monitoring. Obviously that has to occur at the end of the signal chain and really shouldn't be too hard. Just average loudness measure bypassed and non-bypassed, and compensate the non-bypassed accordingly. This should be understood as a monitoring feature for quick A/Bing. I don't know how it works in Ozone, but it really does there - do whatever they do!

[sircuit]

It seems like a pretty handy plugin to me in that it's a great time-and-effort-saving way to make stuff sound better (but not always, depending exactly on what the stuff is).

I can't really see it "a great time-and-effort-saving" as long as I have no clue how much I have to blindly tweak the input (and subsequently adjust the output level) and the intensity and the dry/wet and the bias only to find out that the sound is not improving.

Where I'm much more dubious is that the marketing around this leads you to believe that something amazing and revolutionary is going on when it really isn't, and the occasional suggestion that you can make stuff sound louder without affecting the LUFS which is I think completely wrong.

To that end, I'd definitely be in favour of real volume compensation monitoring. Obviously that has to occur at the end of the signal chain and really shouldn't be too hard. Just average loudness measure bypassed and non-bypassed, and compensate the non-bypassed accordingly. This should be understood as a monitoring feature for quick A/Bing. I don't know how it works in Ozone, but it really does there - do whatever they do!

TDR made an awesome thing in all their plugins called "equal loudness bypass": what they do is permanently analyzing the input and output signals in the background. everytime you adjust a setting, an LED goes red to signal the fact that analysis is being performed and the bypass will not reflect the loudness. However in seconds the LED settles to green and as long as it stays green you can bypass knowing that the inut and output loudness match. If you change parameters the LED goes red for few seconds and so on.

[kiezum]

...which uses technologies also used in some types of facial recognition – assumes that there ARE broad strokes as well as details in the input.

Intensify uses facial recognition? What does the plugin 'see'? Does it recognize patterns in the waveform or the spectral graph?

Why not use the technologies in voice recognition?

[zynaptiq]

Would you say that the best results of the plugin are when Intensity is applied to:

a) Individual tracks
b) Sub-groups (e.g. a grouped collection of drum tracks)
c) Final 2-bus/mixbus

Also would this be applied before the final mastering or as part of the final mastering?

Cheers

It can work for any or all of those.

a) For individual tracks, I would probably i) aim for a lower processing amount except maybe on lead instruments, as similar to the psychological effect that an exciter can have you may end up overdoing it in total, and ii) try to carve out a little more of a window mix situation using the custom bias curve sliders. The guitar example in the trailer for example, I used slightly different settings for the two guitars, which gives them a little better separation. Also, I tend to first set the INTENSITY amount with LVL COMP off, THEN set the channel level into the mix. Chances are that you'll be bringing the mix level down to a lower level than you would have otherwise (similar effect to how a compressor would affect such a scenario).

b) sub groups – good place for the plugin. Here, switching saturation on but not boosting the gain can be a cool way of eating some peaks while bringing up the density much more than the actual gain reduction (from the saturate) would otherwise do. If early in the mixing / writing stage, I'd probably leave LVL COMP off here too, as i) there's no need to adhere to a specific given level, and ii) what sounds good IS good at this point.

c) final mix. Here, I would definitely keep LVL COMP on. Start with Dry/Wet at 100% wet, set INTENSITY and the BIAS stuff up as needed - then, take the dry/wet all the way down and bring it up gradually. This gives you a good "mental reset". Often, I end up going with around 50% wet, so about half the amount I initially dialled in.

Depending on how...hmm...well-behaved your source material is, the saturate may want to come on as a limiter stage. In general though, i would always recommend having a dedicated true peak limiter as the last stage.
I would probably do INTENSITY before any compressors/multibands as you may end up using less of those as the density aspect is something INTENSITY brings to the table.

[zynaptiq]

...which uses technologies also used in some types of facial recognition – assumes that there ARE broad strokes as well as details in the input.

Intensify uses facial recognition? What does the plugin 'see'? Does it recognize patterns in the waveform or the spectral graph?

Why not use the technologies in voice recognition?

What it "sees" are "big structures" and "small structures" in a special type of time-frequency domain representation; similar to a spectral graph in that it has the dimensions time, frequency, level, but not based on an FFT or similar.

We amplify the "Small structures" for the INTENSITY effect – it's similar to some image sharpeners.

[noiseboyuk]

I can't really see it "a great time-and-effort-saving" as long as I have no clue how much I have to blindly tweak the input (and subsequently adjust the output level) and the intensity and the dry/wet and the bias only to find out that the sound is not improving.

Sure, if it's not improving then no. But I've already heard enough examples to deduce that this will work for many scenarios very well. I think inherently it will be less interesting to those who like to get forensic with mixes and masters and have the skillset to go with it, and more appealing to those who either need stuff done in a hurry or lack specialist knowledge / experience.

[martinjuenke]

... and more appealing to those who either need stuff done in a hurry or lack specialist knowledge / experience.

That describes me very well ...

[jbarish]

I think that the requests for volume compensation at the output of the plug would clear up alot of the confusion here. Also an input fader so you can more precisely control how hard you want to drive Intensity would be a big help as well. Please consider adding these as most other plugins have both these days and it would add greater flexibility to your product.

[sircuit]

We amplify the "Small structures" for the INTENSITY effect – it's similar to some image sharpeners.

Isn't the sharpen of an image enhancing the edges and thus making the image "sharp"? In order for a hair to be visible the effect is to darken the hair while simultaneusly lighten the surrounding area. A sharp image hase more separation visually.

If I were to compare sharpness with an audio property that would be the transients. A sharp sound is a a sound with audible transients and clear separation between elements. To me, Intensity is rather blurring things: bringing the background to foreground, reducing the transients of the foreground, basically it blends everything together.

This would be a smart and useful plugin: analyze the sound, fingerprint/identify the instruments and apply spectral processing to enhance the separation between them. That would worth €350.