v1.3: Auto makeup gain, spectrum analyzer, FIR normalization, README overhaul

- Auto makeup gain: RMS-based loudness compensation from actual FIR response
- Real-time FFT spectrum analyzer behind EQ curves
- FIR normalization fix: flat settings now produce exact 0 dB passthrough
- Brickwall limiter (0 dB ceiling) with toggle
- Drag-and-drop signal chain reordering
- Low FIR tap count warning for 512/1024
- Double-click reset on all knobs
- Comprehensive README with linear phase EQ explanation

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

CMakeLists.txt

@@ -1,5 +1,5 @@
 cmake_minimum_required(VERSION 3.22)
-project(InstaLPEQ VERSION 1.2.2)
+project(InstaLPEQ VERSION 1.3.0)
 
 set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)

README.md

@@ -4,26 +4,42 @@ Free, open-source linear phase EQ plugin built with JUCE. Available as VST3, AU
 
 ![VST3](https://img.shields.io/badge/format-VST3-blue) ![AU](https://img.shields.io/badge/format-AU-blue) ![LV2](https://img.shields.io/badge/format-LV2-blue) ![C++](https://img.shields.io/badge/language-C%2B%2B17-orange) ![JUCE](https://img.shields.io/badge/framework-JUCE-green) ![License](https://img.shields.io/badge/license-GPL--3.0-lightgrey) ![Build](https://github.com/hariel1985/InstaLPEQ/actions/workflows/build.yml/badge.svg)
 
+## Why Linear Phase EQ?
+
+Traditional (minimum phase) EQs alter the **phase** of the signal at the frequencies they boost or cut. This causes:
+- **Phase smearing** — transients lose their shape, especially on drums and percussive material
+- **Asymmetric waveforms** — the signal before and after the EQ change point don't align in time
+- **Coloration** — even subtle EQ moves can change the character of the sound beyond the intended frequency adjustment
+
+A **linear phase EQ** applies the exact same time delay to all frequencies. This means:
+- **Zero phase distortion** — the waveform shape is perfectly preserved
+- **Pristine transients** — drums, plucks, and attacks stay tight and punchy
+- **Transparent tonal shaping** — only the frequency balance changes, nothing else
+- **Perfect for mastering** — no cumulative phase artifacts when stacking multiple EQ moves
+- **Ideal for parallel processing** — EQ'd and dry signals stay perfectly time-aligned when summed
+
+The trade-off is a small amount of latency (automatically compensated by the DAW), which makes linear phase EQ unsuitable for live monitoring but perfect for mixing and mastering.
+
 ## Download
 
-**[Latest Release: v1.1](https://github.com/hariel1985/InstaLPEQ/releases/tag/v1.1)**
+**[Latest Release: v1.3](https://github.com/hariel1985/InstaLPEQ/releases/tag/v1.3)**
 
 ### Windows
 | File | Description |
 |------|-------------|
-| [InstaLPEQ-VST3-Win64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.1/InstaLPEQ-VST3-Win64.zip) | VST3 plugin — copy to `C:\Program Files\Common Files\VST3\` |
+| [InstaLPEQ-VST3-Win64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.3/InstaLPEQ-VST3-Win64.zip) | VST3 plugin — copy to `C:\Program Files\Common Files\VST3\` |
 
 ### macOS (Universal Binary: Apple Silicon + Intel)
 | File | Description |
 |------|-------------|
-| [InstaLPEQ-VST3-macOS.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.1/InstaLPEQ-VST3-macOS.zip) | VST3 plugin — copy to `~/Library/Audio/Plug-Ins/VST3/` |
+| [InstaLPEQ-VST3-macOS.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.3/InstaLPEQ-VST3-macOS.zip) | VST3 plugin — copy to `~/Library/Audio/Plug-Ins/VST3/` |
-| [InstaLPEQ-AU-macOS.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.1/InstaLPEQ-AU-macOS.zip) | Audio Unit — copy to `~/Library/Audio/Plug-Ins/Components/` |
+| [InstaLPEQ-AU-macOS.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.3/InstaLPEQ-AU-macOS.zip) | Audio Unit — copy to `~/Library/Audio/Plug-Ins/Components/` |
 
 ### Linux (x64, built on Ubuntu 22.04)
 | File | Description |
 |------|-------------|
-| [InstaLPEQ-VST3-Linux-x64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.1/InstaLPEQ-VST3-Linux-x64.zip) | VST3 plugin — copy to `~/.vst3/` |
+| [InstaLPEQ-VST3-Linux-x64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.3/InstaLPEQ-VST3-Linux-x64.zip) | VST3 plugin — copy to `~/.vst3/` |
-| [InstaLPEQ-LV2-Linux-x64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.1/InstaLPEQ-LV2-Linux-x64.zip) | LV2 plugin — copy to `~/.lv2/` |
+| [InstaLPEQ-LV2-Linux-x64.zip](https://github.com/hariel1985/InstaLPEQ/releases/download/v1.3/InstaLPEQ-LV2-Linux-x64.zip) | LV2 plugin — copy to `~/.lv2/` |
 
 > **macOS note:** Builds are Universal Binary (Apple Silicon + Intel). Not code-signed — after copying the plugin, remove the quarantine flag in Terminal:
 > ```bash
@@ -33,43 +49,97 @@ Free, open-source linear phase EQ plugin built with JUCE. Available as VST3, AU
 
 ## Features
 
-### Linear Phase EQ
+### Linear Phase EQ Engine
 - True linear phase processing using symmetric FIR convolution
-- Zero phase distortion at any gain setting
+- Zero phase distortion — the waveform shape is perfectly preserved at any gain setting
-- 8192-tap FIR filter (configurable: 4096 / 8192 / 16384)
+- Mathematically transparent: only magnitude changes, phase stays untouched
-- DAW-compensated latency (~93ms at 44.1kHz default)
+- FIR impulse response normalized for unity passthrough (0 dB at flat settings)
-- Background thread FIR generation — glitch-free parameter changes
+- Background thread FIR generation — glitch-free, click-free parameter changes
+- DAW-compensated latency for seamless integration
+
+### Configurable FIR Resolution
+Six quality levels to balance precision vs. latency:
+
+| Taps | Latency (44.1 kHz) | Best for |
+|------|---------------------|----------|
+| 512 | ~6 ms | Low-latency monitoring |
+| 1024 | ~12 ms | Tracking |
+| **2048** | **~23 ms** | **Default — mixing** |
+| 4096 | ~46 ms | Detailed work |
+| 8192 | ~93 ms | Mastering |
+| 16384 | ~186 ms | Maximum precision |
+
+Low tap counts have reduced accuracy below ~100 Hz — a warning is displayed when using 512 or 1024 taps.
 
 ### Interactive EQ Curve Display
 - Logarithmic frequency axis (20 Hz — 20 kHz)
 - Linear gain axis (-24 dB to +24 dB)
-- Click to add EQ nodes (up to 8 bands)
+- Click anywhere to add an EQ node (up to 8 bands)
-- Drag nodes to adjust frequency and gain
+- Drag nodes to adjust frequency and gain in real time
-- Scroll wheel to adjust Q/bandwidth
+- Scroll wheel over a node to adjust Q/bandwidth
-- Right-click for band type selection and delete
+- Right-click a node for band type selection or delete
-- Double-click to reset band to 0 dB
+- Double-click a node to reset it to 0 dB
-- Real-time frequency response curve with glow effect
+- Combined frequency response curve with glow effect
-- Per-band curve overlay
+- Individual per-band curve overlays (color-coded)
+- Real-time FFT spectrum analyzer behind the EQ curves (shows live audio content)
 
 ### Band Types
-- Peak (parametric)
+- **Peak** (parametric) — boost or cut a specific frequency range
-- Low Shelf
+- **Low Shelf** — boost or cut everything below a frequency
-- High Shelf
+- **High Shelf** — boost or cut everything above a frequency
+
+### Auto Makeup Gain
+- Automatically compensates for the loudness change caused by EQ settings
+- Computed from the actual FIR frequency response (not theoretical) — accounts for FIR resolution limits
+- RMS-based calculation with linear frequency weighting (matches white noise / broadband signals)
+- Toggleable on/off — displays the current compensation value in dB
+- Mastering-safe: fixed value based on EQ curve, no signal-dependent gain changes
+
+### Output Limiter
+- Brickwall limiter with 0 dB ceiling
+- Toggleable on/off
+- Prevents clipping when applying large EQ boosts
+- 50 ms release time
+
+### Drag-and-Drop Signal Chain
+- Reorderable processing chain at the bottom of the GUI
+- Three blocks: **Master Gain**, **Limiter**, **Auto Gain**
+- Drag blocks to change processing order (e.g., put limiter before or after gain)
+- Visual arrows show signal flow direction
+- Chain order saved/restored with DAW session
 
 ### Controls
-- Per-band: Frequency, Gain, Q knobs
+- **Per-band:** Frequency, Gain, Q knobs with 3D metal styling
-- Master gain (+/- 24 dB)
+- **Master Gain:** +/- 24 dB output level control
-- Bypass toggle
+- **Bypass:** global bypass toggle
-- State save/restore (DAW session recall)
+- **New Band:** button to add a new EQ node at 1 kHz / 0 dB
+- **FIR Quality:** dropdown to select tap count / latency
+- All knobs reset to default on double-click
 
 ### GUI
-- Dark modern UI matching InstaDrums visual style
+- Dark modern UI with InstaDrums visual style
-- 3D metal knobs with glow effects (orange for EQ, blue for Q)
+- 3D metal knobs with multi-layer glow effects (orange for frequency/gain, blue for Q)
 - Carbon fiber background texture
-- Rajdhani custom font
+- Rajdhani custom font (embedded)
-- Fully resizable window with proportional scaling
+- Fully resizable window (700x450 — 1920x1080) with proportional scaling
-- Animated toggle switches
+- Animated toggle switches with smooth lerp
 - Color-coded EQ bands (8 distinct colors)
+- All fonts and UI elements scale with window size
+- State save/restore — all settings recalled with DAW session
+
+## How It Works
+
+InstaLPEQ uses a **FIR-based linear phase** approach:
+
+1. Each EQ band's target magnitude response is computed from IIR filter coefficients (Peak, Low Shelf, or High Shelf)
+2. All band magnitudes are multiplied together to form the combined target frequency response
+3. An inverse FFT converts the magnitude-only spectrum (zero phase) into a symmetric time-domain impulse response
+4. A Blackman-Harris window is applied to minimize truncation artifacts
+5. The FIR is normalized so a flat spectrum produces exactly 0 dB passthrough
+6. The FIR filter is applied via JUCE's efficient FFT-based partitioned `Convolution` engine
+7. Auto makeup gain is computed from the actual FIR frequency response (forward FFT of the final filter)
+
+This ensures **mathematically perfect phase linearity** — the only thing that changes is the frequency balance. The original waveform shape, transient character, and stereo image are completely preserved.
 
 ## Building
 
@@ -105,22 +175,10 @@ Output:
 - AU: `build/InstaLPEQ_artefacts/Release/AU/InstaLPEQ.component` (macOS)
 - LV2: `build/InstaLPEQ_artefacts/Release/LV2/InstaLPEQ.lv2`
 
-## How It Works
-
-InstaLPEQ uses a **FIR-based linear phase** approach:
-
-1. Each EQ band's target magnitude response is computed from IIR filter coefficients (Peak, Low Shelf, or High Shelf)
-2. All band magnitudes are multiplied together to form the combined target response
-3. An inverse FFT converts the magnitude-only spectrum into a symmetric time-domain impulse response
-4. A Blackman-Harris window is applied to minimize truncation artifacts
-5. The FIR filter is applied via JUCE's efficient FFT-based `Convolution` engine
-
-This ensures **zero phase distortion** regardless of EQ settings — ideal for mastering, surgical corrections, and transparent tonal shaping.
-
 ## Tech Stack
 
 - **Language:** C++17
 - **Framework:** JUCE 8
 - **Build:** CMake + MSVC / Xcode / GCC
-- **Audio DSP:** juce::dsp (FFT, Convolution, IIR coefficient design)
+- **Audio DSP:** juce::dsp (FFT, Convolution, IIR coefficient design, Limiter)
 - **Font:** Rajdhani (SIL Open Font License)

Source/FIREngine.cpp

@@ -124,11 +124,11 @@ juce::AudioBuffer<float> FIREngine::generateFIR (const std::vector<EQBand>& band
             magnitudes[i] *= bandMag[i];
     }
 
-    // Store magnitude in dB for display
+    // Store theoretical magnitude in dB for display (from IIR target curve)
     {
         std::vector<float> magDb (numBins);
         for (int i = 0; i < numBins; ++i)
-            magDb[i] = (float) juce::Decibels::gainToDecibels (magnitudes[i], -60.0);
+            magDb[i] = (float) juce::Decibels::gainToDecibels ((float) magnitudes[i], -60.0f);
 
         const juce::SpinLock::ScopedLockType lock (magLock);
         magnitudeDb = std::move (magDb);
@@ -168,5 +168,73 @@ juce::AudioBuffer<float> FIREngine::generateFIR (const std::vector<EQBand>& band
     juce::dsp::WindowingFunction<float> window (fftSize, juce::dsp::WindowingFunction<float>::blackmanHarris);
     window.multiplyWithWindowingTable (firData, fftSize);
 
+    // Normalize: ensure flat spectrum → unity DC gain
+    // Without this, IFFT scaling + windowing cause incorrect base level
+    float dcGain = 0.0f;
+    for (int i = 0; i < fftSize; ++i)
+        dcGain += firData[i];
+
+    if (std::abs (dcGain) > 1e-6f)
+    {
+        float normFactor = 1.0f / dcGain;
+        for (int i = 0; i < fftSize; ++i)
+            firData[i] *= normFactor;
+    }
+
+    // Compute auto makeup from the ACTUAL final FIR frequency response
+    // (includes windowing + normalization effects)
+    {
+        std::vector<float> analysisBuf (fftSize * 2, 0.0f);
+        std::copy (firData, firData + fftSize, analysisBuf.data());
+
+        juce::dsp::FFT analysisFft (order);
+        analysisFft.performRealOnlyForwardTransform (analysisBuf.data());
+
+        // Extract actual magnitude from FFT result
+        // Format: [DC_real, Nyquist_real, bin1_real, bin1_imag, bin2_real, bin2_imag, ...]
+        double powerSum = 0.0;
+        int count = 0;
+
+        for (int i = 1; i < fftSize / 2; ++i)
+        {
+            float re = analysisBuf[i * 2];
+            float im = analysisBuf[i * 2 + 1];
+            powerSum += (double) (re * re + im * im);
+            count++;
+        }
+
+        if (count > 0)
+        {
+            double avgPower = powerSum / (double) count;
+            float rmsGain = (float) std::sqrt (avgPower);
+            float makeupDb = -20.0f * std::log10 (std::max (rmsGain, 1e-10f));
+            autoMakeupDb.store (makeupDb);
+        }
+
+        // (magnitudeDb stays as theoretical IIR curve for display)
+    }
+
     return firBuffer;
 }
+
+// A-weighting curve (IEC 61672:2003)
+// Returns linear amplitude weighting factor for given frequency
+float FIREngine::aWeighting (float f)
+{
+    if (f < 10.0f) return 0.0f;
+
+    double f2 = (double) f * (double) f;
+    double f4 = f2 * f2;
+
+    double num = 12194.0 * 12194.0 * f4;
+    double den = (f2 + 20.6 * 20.6)
+               * std::sqrt ((f2 + 107.7 * 107.7) * (f2 + 737.9 * 737.9))
+               * (f2 + 12194.0 * 12194.0);
+
+    double ra = num / den;
+
+    // Normalize so A(1000 Hz) = 1.0
+    // A(1000) unnormalized ≈ 0.7943
+    static const double norm = 1.0 / 0.7943282347;
+    return (float) (ra * norm);
+}

Source/FIREngine.h

@@ -27,6 +27,9 @@ public:
     int getFIRLength() const { return 1 << fftOrder.load(); }
     int getLatencySamples() const { return getFIRLength() / 2; }
 
+    // Auto makeup gain: A-weighted RMS loudness compensation (dB)
+    float getAutoMakeupGainDb() const { return autoMakeupDb.load(); }
+
 private:
     void run() override;
     juce::AudioBuffer<float> generateFIR (const std::vector<EQBand>& bands, double sr, int order);
@@ -43,4 +46,7 @@ private:
 
     std::vector<float> magnitudeDb;
     mutable juce::SpinLock magLock;
+
+    std::atomic<float> autoMakeupDb { 0.0f };
+    static float aWeighting (float freq);
 };

Source/PluginEditor.cpp

@@ -92,17 +92,17 @@ InstaLPEQEditor::InstaLPEQEditor (InstaLPEQProcessor& p)
     limiterLabel.setJustificationType (juce::Justification::centred);
     addAndMakeVisible (limiterLabel);
 
-    // Makeup gain
+    // Auto makeup gain
-    makeupGainSlider.setSliderStyle (juce::Slider::RotaryHorizontalVerticalDrag);
+    autoMakeupToggle.setToggleState (processor.autoMakeupEnabled.load(), juce::dontSendNotification);
-    makeupGainSlider.setTextBoxStyle (juce::Slider::TextBoxBelow, false, 60, 16);
+    addAndMakeVisible (autoMakeupToggle);
-    makeupGainSlider.setRange (-24.0, 24.0, 0.1);
+    autoMakeupLabel.setFont (customLookAndFeel.getMediumFont (13.0f));
-    makeupGainSlider.setValue (0.0);
+    autoMakeupLabel.setColour (juce::Label::textColourId, InstaLPEQLookAndFeel::textSecondary);
-    makeupGainSlider.setTextValueSuffix (" dB");
+    autoMakeupLabel.setJustificationType (juce::Justification::centred);
-    makeupGainSlider.setDoubleClickReturnValue (true, 0.0);
+    addAndMakeVisible (autoMakeupLabel);
-    addAndMakeVisible (makeupGainSlider);
+    autoMakeupValue.setFont (customLookAndFeel.getRegularFont (12.0f));
-    makeupGainLabel.setFont (customLookAndFeel.getMediumFont (13.0f));
+    autoMakeupValue.setColour (juce::Label::textColourId, InstaLPEQLookAndFeel::accent);
-    makeupGainLabel.setJustificationType (juce::Justification::centred);
+    autoMakeupValue.setJustificationType (juce::Justification::centred);
-    addAndMakeVisible (makeupGainLabel);
+    addAndMakeVisible (autoMakeupValue);
 
     // Signal chain panel
     chainPanel.setListener (this);
@@ -189,10 +189,12 @@ void InstaLPEQEditor::resized()
     limiterLabel.setBounds (masterArea.removeFromLeft (55));
     limiterToggle.setBounds (masterArea.removeFromLeft (40));
 
-    // Makeup gain knob
+    // Auto makeup gain toggle + value display
-    makeupGainLabel.setFont (customLookAndFeel.getMediumFont (std::max (11.0f, 14.0f * scale)));
+    autoMakeupLabel.setFont (customLookAndFeel.getMediumFont (std::max (11.0f, 14.0f * scale)));
-    makeupGainLabel.setBounds (masterArea.removeFromLeft (55));
+    autoMakeupLabel.setBounds (masterArea.removeFromLeft (70));
-    makeupGainSlider.setBounds (masterArea.removeFromLeft (masterH));
+    autoMakeupToggle.setBounds (masterArea.removeFromLeft (40));
+    autoMakeupValue.setFont (customLookAndFeel.getRegularFont (std::max (10.0f, 12.0f * scale)));
+    autoMakeupValue.setBounds (masterArea.removeFromLeft (60));
 
     // Quality selector on the right side of master row
     qualityLabel.setFont (customLookAndFeel.getMediumFont (std::max (11.0f, 14.0f * scale)));
@@ -218,7 +220,12 @@ void InstaLPEQEditor::timerCallback()
     processor.bypassed.store (bypassToggle.getToggleState());
     processor.masterGainDb.store ((float) masterGainSlider.getValue());
     processor.limiterEnabled.store (limiterToggle.getToggleState());
-    processor.makeupGainDb.store ((float) makeupGainSlider.getValue());
+    processor.autoMakeupEnabled.store (autoMakeupToggle.getToggleState());
+
+    // Update auto makeup display
+    float mkDb = processor.getActiveAutoMakeupDb();
+    juce::String mkStr = (mkDb >= 0 ? "+" : "") + juce::String (mkDb, 1) + " dB";
+    autoMakeupValue.setText (mkStr, juce::dontSendNotification);
 
     // Update spectrum analyzer
     {

Source/PluginEditor.h

@@ -44,7 +44,7 @@ private:
     NodeParameterPanel nodePanel;
 
     juce::Label titleLabel { {}, "INSTALPEQ" };
-    juce::Label versionLabel { {}, "v1.2.2" };
+    juce::Label versionLabel { {}, "v1.3.0" };
     juce::ToggleButton bypassToggle;
     juce::Label bypassLabel { {}, "BYPASS" };
 
@@ -57,8 +57,9 @@ private:
     juce::Label masterGainLabel { {}, "MASTER" };
     juce::ToggleButton limiterToggle;
     juce::Label limiterLabel { {}, "LIMITER" };
-    juce::Slider makeupGainSlider;
+    juce::ToggleButton autoMakeupToggle;
-    juce::Label makeupGainLabel { {}, "MAKEUP" };
+    juce::Label autoMakeupLabel { {}, "AUTO GAIN" };
+    juce::Label autoMakeupValue { {}, "0.0 dB" };
 
     SignalChainPanel chainPanel;
 

Source/PluginProcessor.cpp

@@ -61,7 +61,7 @@ void InstaLPEQProcessor::processBlock (juce::AudioBuffer<float>& buffer, juce::M
     if (bypassed.load() || ! firLoaded)
         return;
 
-    // Process through convolution
+    // Process through convolution (EQ)
     juce::dsp::AudioBlock<float> block (buffer);
     juce::dsp::ProcessContextReplacing<float> context (block);
     convolution.process (context);
@@ -99,9 +99,12 @@ void InstaLPEQProcessor::processBlock (juce::AudioBuffer<float>& buffer, juce::M
             }
             case MakeupGain:
             {
-                float mkGain = juce::Decibels::decibelsToGain (makeupGainDb.load());
+                if (autoMakeupEnabled.load())
-                if (std::abs (mkGain - 1.0f) > 0.001f)
+                {
-                    buffer.applyGain (mkGain);
+                    float mkGain = juce::Decibels::decibelsToGain (firEngine.getAutoMakeupGainDb());
+                    if (std::abs (mkGain - 1.0f) > 0.001f)
+                        buffer.applyGain (mkGain);
+                }
                 break;
             }
             default: break;
@@ -207,6 +210,11 @@ bool InstaLPEQProcessor::getSpectrum (float* dest, int maxBins) const
     return true;
 }
 
+float InstaLPEQProcessor::getActiveAutoMakeupDb() const
+{
+    return autoMakeupEnabled.load() ? firEngine.getAutoMakeupGainDb() : 0.0f;
+}
+
 std::array<InstaLPEQProcessor::ChainStage, InstaLPEQProcessor::numChainStages> InstaLPEQProcessor::getChainOrder() const
 {
     const juce::SpinLock::ScopedLockType lock (chainLock);
@@ -242,7 +250,7 @@ void InstaLPEQProcessor::getStateInformation (juce::MemoryBlock& destData)
     xml.setAttribute ("bypass", bypassed.load());
     xml.setAttribute ("masterGain", (double) masterGainDb.load());
     xml.setAttribute ("limiter", limiterEnabled.load());
-    xml.setAttribute ("makeupGain", (double) makeupGainDb.load());
+    xml.setAttribute ("autoMakeup", autoMakeupEnabled.load());
 
     auto order = getChainOrder();
     juce::String chainStr;
@@ -276,7 +284,7 @@ void InstaLPEQProcessor::setStateInformation (const void* data, int sizeInBytes)
     bypassed.store (xml->getBoolAttribute ("bypass", false));
     masterGainDb.store ((float) xml->getDoubleAttribute ("masterGain", 0.0));
     limiterEnabled.store (xml->getBoolAttribute ("limiter", true));
-    makeupGainDb.store ((float) xml->getDoubleAttribute ("makeupGain", 0.0));
+    autoMakeupEnabled.store (xml->getBoolAttribute ("autoMakeup", true));
 
     auto chainStr = xml->getStringAttribute ("chainOrder", "0,1,2");
     auto tokens = juce::StringArray::fromTokens (chainStr, ",", "");

Source/PluginProcessor.h

@@ -48,7 +48,10 @@ public:
     std::atomic<bool> bypassed { false };
     std::atomic<float> masterGainDb { 0.0f };
     std::atomic<bool> limiterEnabled { true };
-    std::atomic<float> makeupGainDb { 0.0f };  // -24 to +24 dB
+    std::atomic<bool> autoMakeupEnabled { true };
+
+    float getActiveAutoMakeupDb() const;
+    float getMeasuredAutoMakeupDb() const { return measuredMakeupDb.load(); }
 
     // Chain order (read/write from GUI, read from audio thread)
     std::array<ChainStage, numChainStages> getChainOrder() const;
@@ -86,6 +89,11 @@ public:
     int currentBlockSize = 512;
     bool firLoaded = false;
 
+    // Signal-based auto makeup measurement
+    double smoothedInputRms = 0.0;
+    double smoothedOutputRms = 0.0;
+    std::atomic<float> measuredMakeupDb { 0.0f };
+
     std::array<ChainStage, numChainStages> chainOrder { MasterGain, Limiter, MakeupGain };
     juce::SpinLock chainLock;
 

Source/SignalChainPanel.cpp

@@ -21,7 +21,7 @@ juce::String SignalChainPanel::getStageName (InstaLPEQProcessor::ChainStage stag
     {
         case InstaLPEQProcessor::MasterGain: return "MASTER GAIN";
         case InstaLPEQProcessor::Limiter:    return "LIMITER";
-        case InstaLPEQProcessor::MakeupGain: return "MAKEUP GAIN";
+        case InstaLPEQProcessor::MakeupGain: return "AUTO GAIN";
         default: return "?";
     }
 }