> Guide authoring of Audio Toolbox plugins (audioPlugin, audioPluginSource) that pass validateAudioPlugin and generate deployable VST/AU code. Use when creating audio effect or generator plugins, writing classdef files inheriting from audioPlugin, or troubleshooting validateAudioPlugin failures.
npx skills add https://github.com/matlab/matlab-agentic-toolkit --skill matlab-write-audio-plugin
audioPlugin or audioPluginSource)validateAudioPlugin or generateAudioPlugin failuresaudioDatastore, audioFeatureExtractor)Every audio plugin is a classdef with %#codegen, public tunable properties, a Constant PluginInterface, and methods process + reset.
classdef MyPlugin < audioPlugin
%#codegen
properties
Gain = 0.5
Cutoff = 1000
end
properties (Access = private)
pSR = 44100
pB = [1 0 0]
pA = [1 0 0]
pState = zeros(2, 2) % (filterOrder, numChannels)
end
properties (Constant)
PluginInterface = audioPluginInterface( ...
audioPluginParameter('Gain', ...
DisplayName='Gain', Label='dB', Mapping={'lin', 0, 1}), ...
audioPluginParameter('Cutoff', ...
DisplayName='Cutoff', Label='Hz', Mapping={'log', 20, 20000}), ...
InputChannels=2, OutputChannels=2)
end
methods
function y = process(plugin, x)
[y, plugin.pState] = filter(plugin.pB, plugin.pA, x, plugin.pState);
y = y * plugin.Gain;
end
function reset(plugin)
plugin.pSR = getSampleRate(plugin);
plugin.pState = zeros(2, 2);
designFilter(plugin);
end
function set.Cutoff(plugin, val)
plugin.Cutoff = val;
designFilter(plugin); %#ok<MCSUP>
end
end
methods (Access = private)
function designFilter(plugin)
wn = plugin.Cutoff / (plugin.pSR / 2);
wn = max(eps, min(wn, 1 - eps));
[plugin.pB, plugin.pA] = butter(2, wn);
end
end
end
Source plugins inherit audioPluginSource, omit InputChannels, and process takes no audio input — use getSamplesPerFrame(plugin) for output frame size.
System Object hybrids (matlab.System & audioPlugin) require (StrictDefaults), isInputSizeMutableImpl returning true, and stepImpl/resetImpl instead of process/reset. Use only when user requests Simulink compatibility.
getSampleRate returns 44100 here; use for initial buffer sizing only.reset — Called when sample rate or frame size changes. Cache getSampleRate(plugin) in pSR. Recompute all SR-dependent values. Call reset() on every sub-object (after setting their sample rate).process — Called per audio frame. Return double output sized [N, numOutputChannels]. Never assign to properties registered in audioPluginParameter — for output meters, keep properties public but unregistered.%#ok<MCSUP> only when the set method actually accesses another property through obj.PropertyName or calls a method that does (e.g., designFilter(plugin) from set.Cutoff). Do not add it preemptively — only suppress warnings that checkcode actually raises on that line.reset cannot reconstruct from parameters and sample rate (e.g., loaded data, handle sub-objects with internal buffers). Not compiled into generated VSTs/standalones — DAW hosts persist parameters via the plugin interface and call reset on restore. See references/advanced-patterns.md for the pattern.All code must be codegen-compatible. validateAudioPlugin sweeps 5 sample rates (8000–192000), frame sizes 2.^(1:13)+1 (max 8193), and all parameter extremes.
Always use built-in functions over hand-implementations. Prefer toolbox functions in this order: Audio Toolbox → DSP System Toolbox → Signal Processing Toolbox → base MATLAB. If unsure whether a codegen-compatible built-in exists for an operation, consult references/available-functions.md before implementing manually.
Pre-allocate all state to maximum needed size. Codegen locks property size from the constructor's last assignment — allocate at the maximum the parameter can reach, then index into the active region at runtime.
Shift with indexed assignment:
buf(1:end-N) = buf(N+1:end); % shift left by N
buf(end-N+1:end) = newData; % fill tail
Never concatenate to shift — [buf(N+1:end); zeros(N,ch)] produces a variable-size result that codegen rejects on assignment to a fixed-size property.
Never use : on the column dimension when assigning to a fixed-size property — codegen treats x(i,:) as variable-size. Index columns explicitly: [x(i,1), x(i,2)].
Initialize arrays that will hold complex values with complex(zeros(...)) — not bare zeros(...). Codegen locks the real/complex attribute from the first assignment. If the array starts real, assigning FFT output into it later fails with "left-hand side constrained to be non-complex."
dsp.AsyncBuffer capacity must exceed the largest single write the plugin will receive. validateAudioPlugin delivers frames up to 8193 samples; account for that plus any overlap when sizing the buffer.
When filter() input comes from a sub-object (e.g., crossoverFilter), codegen cannot propagate the column count — the returned state becomes variable-size. Process channels explicitly:
[low(:,1), plugin.pState(:,1)] = filter(b, a, low(:,1), plugin.pState(:,1));
[low(:,2), plugin.pState(:,2)] = filter(b, a, low(:,2), plugin.pState(:,2));
Use Style='dropdown' for 3+ values, 'vrocker'/'vtoggle' for exactly 2.
When enum values have different character lengths ('On'/'Off', 'Short'/'VeryLong'), prefer a separate int32 enum class file. The framework can handle char padding internally, but an explicit enum class produces cleaner generated code with typed dispatch instead of string comparisons:
% MyMode.m
classdef MyMode < int32
enumeration
Normal (0)
Aggressive (1)
Subtle (2)
end
end
Plugin property: Mode = MyMode.Normal with Mapping={'enum','Normal','Aggressive','Subtle'}.
When computing normalized frequency or delay indices from parameters, clamp the result:
wn = plugin.Cutoff / (plugin.pSR / 2);
wn = max(eps, min(wn, 1 - eps)); % keep in valid (0,1) range for butter/cheby
Clamp delay-line read indices to [1, bufferLength]. This prevents out-of-bounds at extreme sample rates or parameter settings that validateAudioPlugin will exercise.
Every switch that assigns a variable must include otherwise with a safe default — codegen requires all branches to define the same outputs.
Construct with literal arguments in the constructor. In reset, propagate sample rate then reset:
function reset(plugin)
fs = getSampleRate(plugin);
setSampleRate(plugin.pEcho, fs);
reset(plugin.pEcho);
plugin.pCompressor.SampleRate = fs;
reset(plugin.pCompressor);
end
Call sub-plugins as process(plugin.pSub, x). Call System Objects as plugin.pObj(x). Forward parameter changes in set methods.
coder.load('data.mat') in the constructor — baked into the binaryfprintf/disp/plot in if isempty(coder.target)validateAudioPlugin -nomex ClassName % structural + testbench
validateAudioPlugin ClassName % full MEX codegen
generateAudioPlugin ClassName % produce VST/AU binary
After validation passes:
checkcode on all produced .m files — resolve every warning by renaming or restructuring, not by adding %#ok suppressions (except %#ok<MCSUP> in set methods). Re-run checkcode to confirm zero warnings remain.setSampleRate, reset, process a test signal, confirm output matches intent.Default output is VST 2. Use flags to select other formats:
| Flag | Format | Platform |
|------|--------|----------|
| -vst | VST 2 (default) | Windows, macOS |
| -vst3 | VST 3 | Windows, macOS |
| -au | Audio Unit v2 | macOS only |
| -auv3 | Audio Unit v3 | macOS only |
| -exe | Standalone executable | Windows, macOS |
| -juceproject | JUCE project (source code) | All (including Linux) |
When the user requests an output format unavailable on the current platform (e.g., AU on Windows, or any compiled binary on Linux), explain the constraint and suggest the closest alternative (-juceproject on Linux, -vst/-vst3 on Windows instead of AU).
references/available-functions.md — Built-in streaming DSP objects (prefer over manual implementations)references/advanced-patterns.md — AsyncBuffer, modulated delay, save/load, codegen edge casesreferences/parameters.md — Mapping laws, multi-bus I/O, grid layout constraintsreferences/grid-layout.md — Grid layout syntax, row allocation with DisplayNameLocationreferences/deep-learning.md — Neural network inference with coder.loadDeepLearningNetwork----
Copyright 2026 The MathWorks, Inc.
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Improves the quality of images, especially screenshots, by enhancing resolution, sharpness, and clarity. Perfect for preparing images for presentations, documentation, or social media posts.
Downloads videos from YouTube and other platforms for offline viewing, editing, or archival. Handles various formats and quality options.
Lightweight WSI tile extraction and preprocessing. Use for basic slide processing tissue detection, tile extraction, stain normalization for H&E images. Best for simple pipelines, dataset preparation, quick tile-based analysis. For advanced spatial proteomics, multiplexed imaging, or deep learning pipelines use pathml.
Microscopy data management platform. Access images via Python, retrieve datasets, analyze pixels, manage ROIs/annotations, batch processing, for high-content screening and microscopy workflows.
Python library for working with DICOM (Digital Imaging and Communications in Medicine) files. Use this skill when reading, writing, or modifying medical imaging data in DICOM format, extracting pixel data from medical images (CT, MRI, X-ray, ultrasound), anonymizing DICOM files, working with DICOM metadata and tags, converting DICOM images to other formats, handling compressed DICOM data, or processing medical imaging datasets. Applies to tasks involving medical image analysis, PACS systems, radiology workflows, and healthcare imaging applications.
This skill should be used when working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks. Use for text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning models on custom datasets.
Take matlab/matlab-write-audio-plugin from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.