@DanH no it figures it out using advanced AI technology:

numWavetables = fileLength / cycleLength.

- allow CSS & scripted LAF within same component · christophhart/HISE@e459f0b

The open source framework for sample based instruments - - allow CSS & scripted LAF within same component · christophhart/HISE@e459f0b

GitHub (github.com)

Long overdue, but it seems that this was really holding people back from CSS...

These are the rules:

• you can now use local LAF objects with mixed CSS & script functions
• global look and feels can still be only script or CSS but that should be fine as you can always use a local LAF if you need both
• script functions have the precedence, so if a draw function was registered it will always supercede the CSS definition
• the exception is popup menu styling - as soon as you combine script functions & CSS, the popup menu will always be styled with the CSS.

Combobox with scripted draw function & CSS styled popupmenu:

const var laf = Content.createLocalLookAndFeel();

// Combobox is styled by a function
laf.registerFunction("drawComboBox", function(g, obj)
{
	g.setColour(obj.bgColour);
	g.fillRect(obj.area);
	g.setFont("Comic Sans MS", 21);
	g.setColour(Colours.blue);
	g.drawAlignedText(obj.text, obj.area, "centred");
});

// Popup menu is styled by CSS
laf.setInlineStyleSheet("
.popup
{
	background: #333;
}

.popup-item
{
	background: transparent;
	color: #999;
	padding: 10px;
}

.popup-item:hover
{
	background: rgba(255,255,255, 0.2);
}

.popup-item:active
{
	color: white;
	font-weight: bold;
}
");

Content.getComponent("ComboBox1").setLocalLookAndFeel(laf);

so how do you want to merge them? Is Alice 31 or 30?

I would refrain from a recursive logic, looks like you have a fixed depth hierarchy that you can just handcode.

don't use the velocity stacking (I think I've removed those). If you don't need to resample acousting material and already have power of two sized wavetables, just load in a sample that has all the wavetables stitched together, and use the resample mode. Ideally you can pass in the cycle length so the chances of the algorithm messing that up is smaller.

At some point we have to address that problem, I notice more and more posts just pasting the rubbish GPT output which dilutes the prestine information set available here in the forum.

Regarding the problem at hand, a single look at one of the many snippets / code examples from a good source should solve that problem, tuck inline before the function definition and pass it into the call.

@ustk nope once compiled they are a blackbox.

You don‘t need to compile nodes it just makes them faster especially when doing complex stuff and if you need to access them through the API then it‘s definitely a case of leaving them in the interpreted state.

@Oli-Ullmann obj.text

@Oli-Ullmann use the button text as the key in a JSON object with the display string as value then you don‘t need to branch at all.

@Lindon nope it‘s bypassed.

Also it's not stored as plain text, but as compressed zstd blob, but that won't stop the criminal masterminds of this world.

@ustk said in About Hise script seen in plain text in binaries...:

But are the comments exposed as well or are they removed during export?

Nope. But the preprocessor is being processed before the script is embedded, so you can do this:

#on


#if 0
// The PIN for my bank account is 1234.
#endif

and your money will stay safe.

@DabDab but very useful for some!

@d-healey yup good idea.

@orange yup, if you want to use it get in touch with them, I‘m currently helping out a developer with the setup to make the integration more easy, but it‘s a WIP

@JulesV no.

@sinewavekid you need to create multiple networks for each model and then select them from the node via the drop down.

@jmannix do the parameter names of your compiled network contain whitespace characters?

@benzenewings Try the latest commits from this weekend, I just updated macOS to reproduce this and ran into the same error (seems like they've updated the Ruby runtime which broke xcpretty).

I've completely replaced xcpretty with another console formatting app (which is a binary that is now directly included in the HISE source code and not dependent on Ruby) so you shouldn't run into this issue. Still not sure if that was your original problem, but I can now export through the compile project window on the latest macOS stuff without issues.

@griffinboy yes it's included within the RTNeural code folder, but it's only available in the translation unit that compiles the RTNeural code (the hi_tools module).

In order to make it accessible on other translation units and your own nodes you have to add this line to hi_tools/hi_tools.h:

#include "hi_neural/RTNeural/modules/xsimd/xsimd.hpp"

The place doesn't really matter, but somewhere around line 150 would be a good spot.

@hujackus yup, it was missing the doc string in the header file so it wasn't picked up by the API generator. It's fixed now.