VST Plugin Creation from Matlab Script

Hello Gabriele!

I wanna create a VST Plugin for real time audio processing that simulates a shoebox and it auralizes the sound you give it as an input mono or stereo, and with the use of HRTF data and HelperImageSource method , filters/auralizes the output sound. I have my matlab script ready and working , but I am dealing with issues In my conversion of the Matlab script to VST Coding in order to audiotestbench / prototype and generate my audio plugin.

roomDimensions = [25, 20, 17];
receiverCoord = [5, 6, 8];
sourceCoord = [5, 8, 9];
A = [ ... 
    0.10 0.20 0.40 0.60 0.50 0.60; ...
    0.10 0.20 0.40 0.60 0.50 0.60; ...
    0.10 0.20 0.40 0.60 0.50 0.60; ...
    0.10 0.20 0.40 0.60 0.50 0.60; ...
    0.02 0.03 0.03 0.03 0.04 0.07; ...
    0.02 0.03 0.03 0.03 0.04 0.07].';
FVect = [125 250 500 1000 2000 4000];
fs = 48000;
useHRTF = true;
ARIDataset = load("ReferenceHRTF.mat");
hrtfData = permute(ARIDataset.hrtfData, [2, 3, 1]);
sourcePosition = ARIDataset.sourcePosition(:, [1, 2]);
h = HelperImageSource(roomDimensions, receiverCoord, sourceCoord, A, FVect, fs, useHRTF, hrtfData, sourcePosition);
disp(size(h)); % Επιβεβαίωσε τις διαστάσεις του h
% --- Initialize audio input ---
% Load an audio file (replace 'your_audio_file.wav' with your file name)
[audioIn, fs] = audioread('FunkyDrums-44p1-stereo-25secs.mp3');
% Define playback duration in seconds
T = 5; % Set to 5 seconds (or any desired duration)
% Ensure audio duration is sufficient for playback
if length(audioIn) < fix(fs*T)
    error('Audio file is shorter than the requested playback duration T.');
end
% --- Test original audio ---
sound(audioIn(1:fix(fs*T)), fs); % Fix ensures integer sample length
pause(T);
% --- Visualize original audio signal ---
% Create time axis for plotting
time = (0:length(audioIn)-1) / fs;
% Plot original signal
figure;
subplot(2, 1, 1);
plot(time, audioIn);
title('Original Audio Signal');
xlabel('Time (s)');
ylabel('Amplitude');
grid on;
% --- Apply audio filter ---
y1 = filter(h(:, 1), 1, audioIn);
y2 = filter(h(:, 2), 1, audioIn);
y = [y1 y2];
y = y / max(abs(y(:))); % Normalize
% --- Save filtered audio signal ---
outputFileName = 'filtered_audio2.wav'; % Specify the output file name
audiowrite(outputFileName, y, fs); % Save the filtered audio
disp(['Filtered audio saved to ', outputFileName]);
% --- Test filtered audio ---
sound(y(1:round(fs*T)), fs); % Round ensures integer sample length
% --- Visualize filtered audio signal ---
subplot(2, 1, 2);
plot(time, y);
title('Filtered Audio Signal');
xlabel('Time (s)');
ylabel('Amplitude');
grid on;

This is the Matlab initial script code and the code I wrote to convert it to a VST is this below:

classdef RoomAudioSimulator2 < audioPlugin
    % RoomAudioSimulator: A VST plugin simulating room acoustics.
    properties
        roomLength = 25; % Room length
        roomWidth = 20; % Room width
        roomHeight = 17; % Room height
        receiverX = 5; % Receiver X-coordinate
        receiverY = 6; % Receiver Y-coordinate
        receiverZ = 8; % Receiver Z-coordinate
        sourceX = 5; % Source X-coordinate
        sourceY = 8; % Source Y-coordinate
        sourceZ = 9; % Source Z-coordinate
    end
    properties (Constant)
        PluginInterface = audioPluginInterface( ...
            audioPluginParameter('roomLength', ...
            'DisplayName', 'Room Length', ...
            'Mapping', {'lin', 5, 35}), ...
            audioPluginParameter('roomWidth', ...
            'DisplayName', 'Room Width', ...
            'Mapping', {'lin', 5, 35}), ...
            audioPluginParameter('roomHeight', ...
            'DisplayName', 'Room Height', ...
            'Mapping', {'lin', 5, 35}), ...
            audioPluginParameter('receiverX', ...
            'DisplayName', 'Receiver X', ...
            'Mapping', {'lin', 0, 35}), ...
            audioPluginParameter('receiverY', ...
            'DisplayName', 'Receiver Y', ...
            'Mapping', {'lin', 0, 35}), ...
            audioPluginParameter('receiverZ', ...
            'DisplayName', 'Receiver Z', ...
            'Mapping', {'lin', 0, 35}), ...
            audioPluginParameter('sourceX', ...
            'DisplayName', 'Source X', ...
            'Mapping', {'lin', 0, 35}), ...
            audioPluginParameter('sourceY', ...
            'DisplayName', 'Source Y', ...
            'Mapping', {'lin', 0, 35}), ...
            audioPluginParameter('sourceZ', ...
            'DisplayName', 'Source Z', ...
            'Mapping', {'lin', 0, 35}));
    end
    properties (Access = private)
        hrtfData; % HRTF data loaded from ARI dataset
        sourcePosition; % Source positions for HRTF
        h; % Filter coefficients
        fs = 48000; % Sampling rate
        filterState; % Filter state for processing
    end
    methods
        function obj = RoomAudioSimulator2()
            % Load HRTF dataset
            ARIDataset = load("ReferenceHRTF.mat");
            obj.hrtfData = permute(ARIDataset.hrtfData, [2, 3, 1]);
            obj.sourcePosition = ARIDataset.sourcePosition(:, [1, 2]);
            % Initialize filters
            obj.updateFilters();
            obj.reset();
        end
        function audioOut = process(obj, audioIn)
            % Ensure h is initialized
            if isempty(obj.h)
                obj.updateFilters();
            end
            % Handle mono input
            if size(audioIn, 2) == 1
                audioIn = repmat(audioIn, 1, 2);
            elseif size(audioIn, 2) ~= 2
                error('Input audio must have 1 or 2 channels.');
            end
            % Apply filtering with state preservation
            [y1, obj.filterState(:, 1)] = filter(obj.h(:, 1), 1, audioIn(:, 1), obj.filterState(:, 1));
            [y2, obj.filterState(:, 2)] = filter(obj.h(:, 2), 1, audioIn(:, 2), obj.filterState(:, 2));
            audioOut = [y1, y2];
        end
        function reset(obj)
            % Reset filter states
            if isempty(obj.h)
                obj.filterState = zeros(100, 2); % Default size if h is uninitialized
            else
                obj.filterState = zeros(size(obj.h, 1) - 1, 2);
            end
        end
        function update(obj)
            % Called during real-time parameter tuning
            obj.adaptPositions();
            obj.updateFilters();
        end
        function updateFilters(obj)
            % Efficiently update filter coefficients
            roomDimensions = [obj.roomLength, obj.roomWidth, obj.roomHeight];
            receiverCoord = [obj.receiverX, obj.receiverY, obj.receiverZ];
            sourceCoord = [obj.sourceX, obj.sourceY, obj.sourceZ];
            A = [ ... 
                0.10 0.20 0.40 0.60 0.50 0.60; ...
                0.10 0.20 0.40 0.60 0.50 0.60; ...
                0.10 0.20 0.40 0.60 0.50 0.60; ...
                0.10 0.20 0.40 0.60 0.50 0.60; ...
                0.02 0.03 0.03 0.03 0.04 0.07; ...
                0.02 0.03 0.03 0.03 0.04 0.07].';
            FVect = [125, 250, 500, 1000, 2000, 4000];
            obj.h = HelperImageSource(roomDimensions, receiverCoord, ...
                sourceCoord, A, FVect, obj.fs, ...
                true, obj.hrtfData, obj.sourcePosition);
            obj.reset(); % Ensure state is reinitialized
        end
        function adaptPositions(obj)
            % Ensure receiver and source positions stay inside room dimensions
            obj.receiverX = min(max(round(obj.receiverX), 0), obj.roomLength);
            obj.receiverY = min(max(round(obj.receiverY), 0), obj.roomWidth);
            obj.receiverZ = min(max(round(obj.receiverZ), 0), obj.roomHeight);
            obj.sourceX = min(max(round(obj.sourceX), 0), obj.roomLength);
            obj.sourceY = min(max(round(obj.sourceY), 0), obj.roomWidth);
            obj.sourceZ = min(max(round(obj.sourceZ), 0), obj.roomHeight);
        end
    end
end

It's working with no errors on AudioTestBench , but it's for sure wrong because whenever I change the tuneable parameters ( RoomDimensions , source-receiver positioning ) the output sound remains the same , so something is wrong. Another issue is that , I cant seem to make it work like a real-time VST Plugin. It would be so helpful to give me some feedback and help me build this!

Thanks in Advance!

Noulis Dimitrios , Electrical & Electronics Engineer in UniWA, Athens, Greece.

Answer 1

[jibrahim]

Hi DIMITRIOS,

I have a few high-level remarks:

1) On the issue of not hearing any changes when you tune source position and other properties: In an audio plugin, you typically react to a change in the property value in a set method (for example, set.receiverX). In that set method, you should update the room impulse response based on the new parameters. Alternatively, you can manage a logical property (for example, needUpdate), which you set to true in set.receiverX and other such set method. Then, in process, you can re-design the filter if need be:

function audioOut = process(obj, audioIn)
if obj.needUpdate
    obj.updateFilters();
end
...
end

2) There are other considerations when authoring a plugin. If your goal is to eventually generate a plugin (.VST, .AU, etc), then the MATLAB code in your plugin must pass validateAudioPlugin, which your plugin does not pass yet, partly because your MATLAB code does not support code generation.

Moreover, for the plugin to run efficiently in run time, the computation of the impulse response must be fast and efficient enough, otherwise you will epxerience drops whenever you tune parameters. I suspect you are suing HelperImageSource from a shipping example in Audio Toolbox. That helper function is not necessarly suitable for such real-time scenarios.