@newgameplusinc/odyssey-audio-video-sdk-dev 1.0.62 → 1.0.63

package/dist/SpatialAudioManager.d.ts

@@ -112,6 +112,13 @@ export declare class SpatialAudioManager extends EventManager {
      * @param muted True to mute, false to unmute
      */
     setParticipantMuted(participantId: string, muted: boolean): void;
+    /**
+     * Update listener position and orientation
+     * The \"listener\" is YOU - where your ears/head are positioned
+     *
+     * @param position Your HEAD position (camera position), not body position!
+     * @param orientation Which way your head is facing (forward and up vectors)
+     */
     setListenerPosition(position: Position, orientation: {
         forwardX: number;
         forwardY: number;
@@ -122,7 +129,10 @@ export declare class SpatialAudioManager extends EventManager {
     }): void;
     /**
      * Update listener orientation from LSD camera direction
-     * @param cameraPos Camera position in world space
+     * IMPORTANT: Uses CAMERA position (head) as listener, not body position!
+     *
+     * @param listenerPos Player body position (for reference, not used as listener)
+     * @param cameraPos Camera/HEAD position - THIS is the actual listener position for audio
      * @param lookAtPos Look-at position (where camera is pointing)
      */
     setListenerFromLSD(listenerPos: Position, cameraPos: Position, lookAtPos: Position): void;

package/dist/SpatialAudioManager.js

@@ -142,13 +142,14 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         }
         // Create BiquadFilter nodes for static/noise reduction
         // Based on: https://tagdiwalaviral.medium.com/struggles-of-noise-reduction-in-rtc-part-2-2526f8179442
-        // HIGHPASS FILTER: Remove low-frequency rumble (< 80Hz)
+        // HIGHPASS FILTER: Remove low-frequency rumble and plosives (< 80Hz)
         // Human voice fundamental: 80-300Hz, harmonics: 300Hz-8kHz
         // This cuts BELOW voice range while preserving full voice spectrum
+        // Lower Q for gentler slope = less phase distortion = fewer onset artifacts
         const highpassFilter = this.audioContext.createBiquadFilter();
         highpassFilter.type = "highpass";
         highpassFilter.frequency.value = 80; // Cut frequencies below 80Hz (removes rumble/pops)
-        highpassFilter.Q.value = 1.0; // Quality factor
+        highpassFilter.Q.value = 0.707; // Butterworth response (gentler, reduces plosives)
         // LOWPASS FILTER: Remove high-frequency hiss (> 8000Hz)
         // Voice harmonics extend to ~8kHz - this preserves full voice richness
         // while removing digital artifacts and hiss ABOVE useful voice range
@@ -156,14 +157,14 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         lowpassFilter.type = "lowpass";
         lowpassFilter.frequency.value = 8000; // Cut frequencies above 8kHz (preserves voice harmonics)
         lowpassFilter.Q.value = 1.0; // Quality factor
-        // VOICE BAND EMPHASIS: Boost 80-300Hz fundamental range for clarity
+        // VOICE BAND EMPHASIS: Subtle boost 80-300Hz fundamental range for clarity
         // This emphasizes the base pitch without affecting harmonics
-        // Helps reduce the "pop" when someone starts speaking
+        // Reduced gain to prevent onset artifacts ("mic pop" when speaking starts)
         const voiceBandFilter = this.audioContext.createBiquadFilter();
         voiceBandFilter.type = "peaking";
         voiceBandFilter.frequency.value = 180; // Center of voice fundamental (80-300Hz)
-        voiceBandFilter.Q.value = 1.5; // Moderate width (~100-260Hz affected)
-        voiceBandFilter.gain.value = 2; // +2dB boost for clarity
+        voiceBandFilter.Q.value = 0.8; // Wider, gentler curve (reduces artifacts)
+        voiceBandFilter.gain.value = 1; // +1dB subtle boost (was 2dB - too aggressive)
         const dynamicLowpass = this.audioContext.createBiquadFilter();
         dynamicLowpass.type = "lowpass";
         dynamicLowpass.frequency.value = 7500; // Fixed for all angles
@@ -181,13 +182,24 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         panner.coneOuterGain = 0.3; // Some sound even outside cone
         // Configure gain for individual participant volume control
         gain.gain.value = 1.0;
+        // ADD COMPRESSOR: Prevents sudden peaks and "pops" when speaking starts
+        // This is KEY to eliminating onset artifacts
+        const participantCompressor = this.audioContext.createDynamicsCompressor();
+        participantCompressor.threshold.value = -24; // Start compressing at -24dB
+        participantCompressor.knee.value = 10; // Smooth knee for natural sound
+        participantCompressor.ratio.value = 3; // 3:1 compression ratio
+        participantCompressor.attack.value = 0.003; // 3ms fast attack for transients
+        participantCompressor.release.value = 0.15; // 150ms release for natural decay
         let currentNode = source;
+        // First apply compressor to tame initial transients (CRITICAL for preventing pops)
+        currentNode.connect(participantCompressor);
+        currentNode = participantCompressor;
         if (denoiseNode) {
             currentNode.connect(denoiseNode);
             currentNode = denoiseNode;
         }
         // Audio chain with voice optimization filters
-        // Chain: source -> [denoise] -> highpass -> voiceBand -> lowpass -> dynamicLowpass -> proximityGain -> panner -> analyser -> gain -> masterGain
+        // Chain: source -> compressor -> [denoise] -> highpass -> voiceBand -> lowpass -> dynamicLowpass -> proximityGain -> panner -> analyser -> gain -> masterGain
         currentNode.connect(highpassFilter);
         highpassFilter.connect(voiceBandFilter);
         voiceBandFilter.connect(lowpassFilter);
@@ -211,6 +223,7 @@ class SpatialAudioManager extends EventManager_1.EventManager {
             analyser,
             gain,
             proximityGain,
+            compressor: participantCompressor,
             highpassFilter,
             lowpassFilter,
             voiceBandFilter,
@@ -301,10 +314,10 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         const nodes = this.participantNodes.get(participantId);
         if (nodes?.panner) {
             const normalizedPosition = this.normalizePositionUnits(position);
-            const listenerPos = this.listenerPosition;
-            // Calculate distance (in meters)
+            const listenerPos = this.listenerPosition; // This is HEAD position (from setListenerFromLSD)
+            // Calculate distance (in meters) - from HEAD to source
             const distance = this.getDistanceBetween(listenerPos, normalizedPosition);
-            // Calculate angle between listener and source
+            // Calculate angle between listener HEAD and source
             const angle = this.calculateAngle(listenerPos, normalizedPosition, this.listenerDirection.forward);
             // Calculate stereo panning based on angle
             const panning = this.calculatePanning(angle);
@@ -314,10 +327,14 @@ class SpatialAudioManager extends EventManager_1.EventManager {
             this.applyStereoPanning(participantId, panning);
             // Apply gain with smooth transition to reduce clicking/popping
             const gainValue = gain / 100; // Convert to 0-1 range
-            nodes.gain.gain.setTargetAtTime(gainValue, this.audioContext.currentTime, 0.05 // Smooth transition over 50ms to reduce clicking
-            );
-            // Apply proximity gain for additional distance-based attenuation
-            nodes.proximityGain.gain.setTargetAtTime(gainValue, this.audioContext.currentTime, 0.05);
+            // Use exponentialRampToValueAtTime for smoother, more natural transitions
+            // This prevents the "pop" when someone starts speaking
+            const currentTime = this.audioContext.currentTime;
+            const rampTime = 0.08; // 80ms smooth ramp (was 50ms - increased for gentler onset)
+            // Ensure we never ramp to exactly 0 (causes issues)
+            const targetGain = Math.max(0.001, gainValue);
+            nodes.gain.gain.setTargetAtTime(targetGain, currentTime, rampTime);
+            nodes.proximityGain.gain.setTargetAtTime(targetGain, currentTime, rampTime);
             // Update 3D position for PannerNode (still used for vertical positioning)
             nodes.panner.positionY.setValueAtTime(normalizedPosition.y, this.audioContext.currentTime);
             nodes.panner.positionZ.setValueAtTime(normalizedPosition.z, this.audioContext.currentTime);
@@ -348,17 +365,28 @@ class SpatialAudioManager extends EventManager_1.EventManager {
             nodes.gain.gain.setValueAtTime(muted ? 0 : 1, this.audioContext.currentTime);
         }
     }
+    /**
+     * Update listener position and orientation
+     * The \"listener\" is YOU - where your ears/head are positioned
+     *
+     * @param position Your HEAD position (camera position), not body position!
+     * @param orientation Which way your head is facing (forward and up vectors)
+     */
     setListenerPosition(position, orientation) {
         const normalizedPosition = this.normalizePositionUnits(position);
         this.applyListenerTransform(normalizedPosition, orientation);
     }
     /**
      * Update listener orientation from LSD camera direction
-     * @param cameraPos Camera position in world space
+     * IMPORTANT: Uses CAMERA position (head) as listener, not body position!
+     *
+     * @param listenerPos Player body position (for reference, not used as listener)
+     * @param cameraPos Camera/HEAD position - THIS is the actual listener position for audio
      * @param lookAtPos Look-at position (where camera is pointing)
      */
     setListenerFromLSD(listenerPos, cameraPos, lookAtPos) {
-        const normalizedListener = this.normalizePositionUnits(listenerPos);
+        // USE CAMERA POSITION AS LISTENER (head position, not body!)
+        const normalizedListener = this.normalizePositionUnits(cameraPos); // ✅ Changed from listenerPos
         const normalizedCamera = this.normalizePositionUnits(cameraPos);
         const normalizedLookAt = this.normalizePositionUnits(lookAtPos);
         // Calculate forward vector (from camera to look-at point)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@newgameplusinc/odyssey-audio-video-sdk-dev",
-  "version": "1.0.62",
+  "version": "1.0.63",
   "description": "Odyssey Spatial Audio & Video SDK using MediaSoup for real-time communication",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",