feat(indicators): refine audio response

iOS/KeyVox Keyboard/Core/Dictation/AudioIndicatorDriver.swift

@@ -43,7 +43,8 @@ final class AudioIndicatorDriver {
         static let speakingMeterPollInterval: CFTimeInterval = 1.0 / 60.0
         static let sampleFreshnessWindow: TimeInterval = 0.35
         static let speakingSampleFreshnessWindow: TimeInterval = 0.35
-        static let listeningSmoothingRate: CGFloat = 16
+        static let listeningAttackSmoothingRate: CGFloat = 25
+        static let listeningDecaySmoothingRate: CGFloat = 8
         static let speakingAttackSmoothingRate: CGFloat = 34
         static let speakingDecaySmoothingRate: CGFloat = 18
     }
@@ -104,7 +105,9 @@ final class AudioIndicatorDriver {
                 ? Metrics.speakingAttackSmoothingRate
                 : Metrics.speakingDecaySmoothingRate
         } else {
-            smoothingRate = Metrics.listeningSmoothingRate
+            smoothingRate = targetLevel > displayedLevel
+                ? Metrics.listeningAttackSmoothingRate
+                : Metrics.listeningDecaySmoothingRate
         }
 
         let smoothing = CGFloat(min(delta * Double(smoothingRate), 1))

iOS/KeyVox Keyboard/Views/Components/KeyboardLogoBarView.swift

@@ -443,26 +443,21 @@ final class KeyboardLogoBarView: UIControl {
             return flatHeight
         }
 
-        if timelineState.signalState == .inactive {
-            return flatHeight
-        }
-
-        if timelineState.signalState == .lowActivity {
-            let quietWaveOffset = timelineState.lowActivityPhase + Double(index) * 0.8
-            let quietRipple = (sin(quietWaveOffset) * 0.5) + 0.5
-            let wiggleOffset = (timelineState.lowActivityPhase * 0.9) + Double(index) * 1.35
-            let ambientWiggle = (sin(wiggleOffset) * 0.5) + 0.5
-            let quietLevel = min(max(timelineState.displayedLevel / 0.14, 0), 1)
-            return flatHeight
-                + (1.2 * scale)
-                + (CGFloat(quietLevel) * (0.8 * scale))
-                + (CGFloat(ambientWiggle) * (0.9 * scale))
-                + (CGFloat(quietRipple) * (2.0 * scale))
-        }
+        let quietWaveOffset = timelineState.lowActivityPhase + Double(index) * 0.8
+        let quietRipple = (sin(quietWaveOffset) * 0.5) + 0.5
+        let wiggleOffset = (timelineState.lowActivityPhase * 0.9) + Double(index) * 1.35
+        let ambientWiggle = (sin(wiggleOffset) * 0.5) + 0.5
+        let quietLevel = min(max(timelineState.displayedLevel / 0.14, 0), 1)
+        let rippleHeight = flatHeight
+            + (1.2 * scale)
+            + (CGFloat(quietLevel) * (0.8 * scale))
+            + (CGFloat(ambientWiggle) * (0.9 * scale))
+            + (CGFloat(quietRipple) * (2.0 * scale))
 
         let multipliers: [CGFloat] = [0.4, 0.7, 1.0, 0.7, 0.4]
         let dynamicHeight = timelineState.displayedLevel * multipliers[index] * maxHeight
-        return max(minHeight, dynamicHeight)
+        let audioHeight = max(minHeight, dynamicHeight)
+        return max(rippleHeight, audioHeight)
     }
 
     private func pixelAligned(_ rect: CGRect) -> CGRect {

iOS/KeyVox iOS/Core/Audio/AudioRecorder+Streaming.swift

@@ -25,6 +25,8 @@ final class AudioCaptureAccumulator {
     private var lastNonDeadSignalTime: Date = .distantPast
     private var lastVisualActiveSignalTime: Date = .distantPast
     private var lastMeaningfulSpeechTime: Date = .distantPast
+    private var visualMeterPreviousInput: Float = 0
+    private var visualMeterPreviousOutput: Float = 0
     private var currentActiveSignalRunDuration: TimeInterval = 0
     private var maxActiveSignalRunDuration: TimeInterval = 0
     private var hadNonDeadSignal = false
@@ -36,6 +38,8 @@ final class AudioCaptureAccumulator {
             lastNonDeadSignalTime = .distantPast
             lastVisualActiveSignalTime = .distantPast
             lastMeaningfulSpeechTime = .distantPast
+            visualMeterPreviousInput = 0
+            visualMeterPreviousOutput = 0
             currentActiveSignalRunDuration = 0
             maxActiveSignalRunDuration = 0
             hadNonDeadSignal = false
@@ -113,19 +117,35 @@ final class AudioCaptureAccumulator {
             let frames = Array(UnsafeBufferPointer(start: floatData[0], count: frameCount))
             audioData.append(contentsOf: frames)
 
+            let visualMeterHighPassCutoffFrequency: Float = 120
+            let sampleInterval = 1 / Float(outputFormat.sampleRate)
+            let highPassTimeConstant = 1 / (2 * Float.pi * visualMeterHighPassCutoffFrequency)
+            let highPassCoefficient = highPassTimeConstant / (highPassTimeConstant + sampleInterval)
+
             var sumSquares: Float = 0
+            var visualSumSquares: Float = 0
             var peak: Float = 0
+            var previousVisualInput = visualMeterPreviousInput
+            var previousVisualOutput = visualMeterPreviousOutput
             for frame in frames {
                 sumSquares += frame * frame
+
+                let visualFrame = highPassCoefficient * (previousVisualOutput + frame - previousVisualInput)
+                visualSumSquares += visualFrame * visualFrame
+                previousVisualInput = frame
+                previousVisualOutput = visualFrame
+
                 let magnitude = abs(frame)
                 if magnitude > peak {
                     peak = magnitude
                 }
             }
+            visualMeterPreviousInput = previousVisualInput
+            visualMeterPreviousOutput = previousVisualOutput
 
             let rms = sqrt(sumSquares / Float(frameCount))
+            let visualMeterRMS = sqrt(visualSumSquares / Float(frameCount))
             let frameDuration = TimeInterval(Double(convertedBuffer.frameLength) / outputFormat.sampleRate)
-            let level = min(max(sqrt(rms) * 2.5, 0), 1)
 
             let now = Date()
             if peak > deadSignalPeakThreshold {
@@ -143,7 +163,11 @@ final class AudioCaptureAccumulator {
             }
 
             let visualActiveThreshold = activeSignalRMSThreshold * visualActiveSignalThresholdMultiplier
-            if rms > visualActiveThreshold {
+            // Visual meter scaling only. This does not modify captured audio samples.
+            let visualRMS = max(visualMeterRMS - visualActiveThreshold, 0)
+            let level = min(sqrt(visualRMS) * 2.7, 1)
+
+            if visualMeterRMS > visualActiveThreshold {
                 lastVisualActiveSignalTime = now
             }
 

macOS/Core/Audio/AudioRecorder+Session.swift

@@ -67,6 +67,8 @@ extension AudioRecorder {
 
         lastNonDeadSignalTime = Date.distantPast
         lastVisualActiveSignalTime = Date.distantPast
+        visualMeterPreviousInput = 0
+        visualMeterPreviousOutput = 0
         currentActiveSignalRunDuration = 0
         maxActiveSignalRunDuration = 0
         lastCaptureHadNonDeadSignal = false

macOS/Core/Audio/AudioRecorder+Streaming.swift

@@ -57,24 +57,36 @@ extension AudioRecorder: AVCaptureAudioDataOutputSampleBufferDelegate {
             audioData.append(contentsOf: frames)
         }
 
-        // Calculate RMS for UI visualization.
+        let sampleInterval = 1 / Float(outputFormat.sampleRate)
+        let highPassTimeConstant = 1 / (2 * Float.pi * visualMeterHighPassCutoffFrequency)
+        let highPassCoefficient = highPassTimeConstant / (highPassTimeConstant + sampleInterval)
+
+        // Calculate raw capture RMS and a high-passed RMS for UI visualization.
         var sum: Float = 0
+        var visualSum: Float = 0
         var peak: Float = 0
+        var previousVisualInput = visualMeterPreviousInput
+        var previousVisualOutput = visualMeterPreviousOutput
         for frame in frames {
             sum += frame * frame
+
+            let visualFrame = highPassCoefficient * (previousVisualOutput + frame - previousVisualInput)
+            visualSum += visualFrame * visualFrame
+            previousVisualInput = frame
+            previousVisualOutput = visualFrame
+
             let magnitude = abs(frame)
             if magnitude > peak {
                 peak = magnitude
             }
         }
+        visualMeterPreviousInput = previousVisualInput
+        visualMeterPreviousOutput = previousVisualOutput
 
         let rms = sqrt(sum / Float(frames.count))
+        let visualMeterRMS = sqrt(visualSum / Float(frames.count))
         let frameDuration = TimeInterval(Double(convertedBuffer.frameLength) / outputFormat.sampleRate)
 
-        // Visual meter scaling only. This does not modify captured audio samples.
-        // Keep boosted UI response so waveform movement remains readable.
-        let level = min(max(sqrt(rms) * 3.5, 0.0), 1.0)
-
         let now = Date()
         if peak > deadSignalPeakThreshold {
             lastNonDeadSignalTime = now
@@ -90,7 +102,12 @@ extension AudioRecorder: AVCaptureAudioDataOutputSampleBufferDelegate {
         }
 
         let visualActiveThreshold = sessionActiveSignalRMSThreshold * visualActiveSignalThresholdMultiplier
-        if rms > visualActiveThreshold {
+        // Visual meter scaling only. This does not modify captured audio samples.
+        // Remove low-level background noise before boosting the remaining signal.
+        let visualRMS = max(visualMeterRMS - visualActiveThreshold, 0)
+        let level = min(sqrt(visualRMS) * 5.5, 1.0)
+
+        if visualMeterRMS > visualActiveThreshold {
             lastVisualActiveSignalTime = now
         }
 

macOS/Core/Audio/AudioRecorder.swift

@@ -51,10 +51,13 @@ class AudioRecorder: NSObject, ObservableObject {
     var sessionInputVolumeScalar: Float = AudioSilenceGatePolicy.defaultInputVolumeScalar
     var sessionThresholdScale: Float = 1.0
     let deadStateHoldDuration: TimeInterval = 0.35
+    let visualMeterHighPassCutoffFrequency: Float = 120
     let visualActiveStateHoldDuration: TimeInterval = 0.16
     let visualActiveSignalThresholdMultiplier: Float = 1.85
     var lastNonDeadSignalTime: Date = Date.distantPast
     var lastVisualActiveSignalTime: Date = Date.distantPast
+    var visualMeterPreviousInput: Float = 0
+    var visualMeterPreviousOutput: Float = 0
     var currentActiveSignalRunDuration: TimeInterval = 0
     var maxActiveSignalRunDuration: TimeInterval = 0
     var lastCaptureHadNonDeadSignal: Bool = false

macOS/Core/Overlay/AudioIndicatorDriver.swift

@@ -47,7 +47,8 @@ final class AudioIndicatorDriver: ObservableObject {
         static let timerInterval: TimeInterval = AudioIndicatorDriver.animationFrameDuration
         static let meterPollInterval: TimeInterval = 1.0 / 30.0
         static let sampleFreshnessWindow: TimeInterval = 0.35
-        static let smoothingRate: CGFloat = 16
+        static let attackSmoothingRate: CGFloat = 15
+        static let decaySmoothingRate: CGFloat = 5
     }
 
     var sampleProvider: (() -> AudioIndicatorSample?)?
@@ -114,7 +115,10 @@ final class AudioIndicatorDriver: ObservableObject {
 
         refreshSampleIfNeeded(at: timestamp, sample: sample)
 
-        let smoothing = min(delta * Metrics.smoothingRate, 1)
+        let smoothingRate = targetLevel > displayedLevel
+            ? Metrics.attackSmoothingRate
+            : Metrics.decaySmoothingRate
+        let smoothing = min(delta * smoothingRate, 1)
         displayedLevel += (targetLevel - displayedLevel) * smoothing
 
         publishTimelineState(at: timestamp)

macOS/Views/Components/LogoBarView.swift

@@ -162,7 +162,7 @@ private struct IndicatorLogoView: View {
         if phase == .processing {
             return .processing
         }
-        if phase == .listening && timelineState.signalState == .lowActivity {
+        if phase == .listening {
             return .lowActivity
         }
         return nil
@@ -223,10 +223,16 @@ private struct TimelineIndicatorBarsView: View {
             for index in 0..<5 {
                 let barHeight: CGFloat = switch mode {
                 case .lowActivity:
-                    lowActivityBarHeight(
-                        index: index,
-                        phase: phase,
-                        displayedLevel: displayedLevel
+                    max(
+                        lowActivityBarHeight(
+                            index: index,
+                            phase: phase,
+                            displayedLevel: displayedLevel
+                        ),
+                        listeningAudioBarHeight(
+                            index: index,
+                            displayedLevel: displayedLevel
+                        )
                     )
                 case .processing:
                     processingBarHeight(index: index, phase: phase)
@@ -300,9 +306,7 @@ private struct ReactiveIndicatorSegmentView: View {
     }
 
     private var height: CGFloat {
-        let minHeight: CGFloat = isDevModeOversized ? 18 : 6
         let flatHeight: CGFloat = isDevModeOversized ? 9 : 3
-        let maxHeight: CGFloat = isDevModeOversized ? 170 : 30
 
         guard phase == .listening else {
             return flatHeight
@@ -320,12 +324,21 @@ private struct ReactiveIndicatorSegmentView: View {
             )
         }
 
-        let multipliers: [CGFloat] = [0.4, 0.7, 1.0, 0.7, 0.4]
-        let dynamicHeight = timelineState.displayedLevel * multipliers[index] * maxHeight
-        return max(minHeight, dynamicHeight)
+        return listeningAudioBarHeight(
+            index: index,
+            displayedLevel: timelineState.displayedLevel
+        )
     }
 }
 
+private func listeningAudioBarHeight(index: Int, displayedLevel: CGFloat) -> CGFloat {
+    let minHeight: CGFloat = isDevModeOversized ? 18 : 6
+    let maxHeight: CGFloat = isDevModeOversized ? 170 : 30
+    let multipliers: [CGFloat] = [0.4, 0.7, 1.0, 0.7, 0.4]
+    let dynamicHeight = displayedLevel * multipliers[index] * maxHeight
+    return max(minHeight, dynamicHeight)
+}
+
 private func processingBarHeight(index: Int, phase: Double) -> CGFloat {
     let flatHeight: CGFloat = isDevModeOversized ? 9 : 3
     let waveOffset = phase + Double(index) * 0.8