@newgameplusinc/odyssey-audio-video-sdk-dev 1.0.20 → 1.0.22

package/README.md

@@ -79,20 +79,32 @@ sdk.setListenerFromLSD(listenerPos, cameraPos, lookAtPos);
 - **Coordinate normalization** – Unreal sends centimeters; `SpatialAudioManager` auto-detects large values and converts to meters once.
 - **Orientation math** – `setListenerFromLSD()` builds forward/right/up vectors from camera/LookAt to keep the listener aligned with head movement.
 - **Dynamic distance gain** – `updateSpatialAudio()` measures distance from listener → source and applies a smooth rolloff curve, so distant avatars fade to silence.
-- **Noise handling** – the AudioWorklet denoiser now runs an adaptive multi-band gate (per W3C AudioWorklet guidance) before the high/low-pass filters, stripping constant HVAC/fan noise even when the speaker is close.
+- **Noise handling** – the AudioWorklet denoiser now runs an adaptive multi-band gate (per W3C AudioWorklet guidance) before the high/low-pass filters, stripping constant HVAC/fan noise even when the speaker is close. A newly added silence gate mutes tracks entirely after ~250 ms of sub-noise-floor energy, eliminating hiss during dead air without touching spatial cues.
+
+#### Noise-Cancellation Stack (What’s Included)
+1. **Adaptive denoiser worklet** – learns each participant’s noise floor in real time, then applies a multi-band downward expander plus dynamic low/high-pass shaping.
+2. **Optional voice enhancement** – autocorrelation-derived confidence (inspired by the tuner article) can raise the reduction floor when speech is present to keep vocals bright.
+3. **Silence gate** – if energy stays below `silenceFloor` for a configurable hold window, the track ramps to true silence, removing hiss when nobody speaks.
+4. **Classic filters** – fixed high-pass/low-pass filters shave off rumble and hiss before signals reach the HRTF panner.
+
+These layers run entirely in Web Audio, so you can ship “AirPods-style” background rejection in any browser without native code.
 	```ts
 	const sdk = new OdysseySpatialComms(serverUrl, {
 		denoiser: {
-			threshold: 0.0085,
-			maxReduction: 0.94,
-			hissCut: 0.7,
-			holdMs: 180,
+			threshold: 0.0082,
+			maxReduction: 0.85,
+			hissCut: 0.5,
+			holdMs: 190,
 			voiceBoost: 0.7,
 			voiceSensitivity: 0.3,
+			voiceEnhancement: true,
+			silenceFloor: 0.0006,
+			silenceHoldMs: 420,
+			silenceReleaseMs: 260,
 		},
 	});
 	```
-	Tweak these knobs if you need even more “AirPods Pro” style isolation.
+	Voice enhancement (autocorrelation-based speech detection) is **off by default** to keep the gate extra quiet; enable it when you want brighter close-talk voicing. Pair this with the new, gentler proximity curve to keep far-away participants airy instead of “all in one room.”
 
 #### How Spatial Audio Is Built
 1. **Telemetry ingestion** – each LSD packet is passed through `setListenerFromLSD(listenerPos, cameraPos, lookAtPos)` so the Web Audio listener matches the player’s real head/camera pose.

package/dist/SpatialAudioManager.d.ts

@@ -19,6 +19,10 @@ type DenoiserOptions = {
     learnRate?: number;
     voiceBoost?: number;
     voiceSensitivity?: number;
+    voiceEnhancement?: boolean;
+    silenceFloor?: number;
+    silenceHoldMs?: number;
+    silenceReleaseMs?: number;
 };
 export type SpatialAudioOptions = {
     distance?: SpatialAudioDistanceConfig;

package/dist/SpatialAudioManager.js

@@ -90,6 +90,10 @@ class SpatialAudioManager extends EventManager_1.EventManager {
                         learnRate: this.options.denoiser?.learnRate,
                         voiceBoost: this.options.denoiser?.voiceBoost,
                         voiceSensitivity: this.options.denoiser?.voiceSensitivity,
+                        voiceEnhancement: this.options.denoiser?.voiceEnhancement,
+                        silenceFloor: this.options.denoiser?.silenceFloor,
+                        silenceHoldMs: this.options.denoiser?.silenceHoldMs,
+                        silenceReleaseMs: this.options.denoiser?.silenceReleaseMs,
                     },
                 });
             }
@@ -506,8 +510,8 @@ class SpatialAudioManager extends EventManager_1.EventManager {
             return;
         }
         const clarityScore = this.calculateClarityScore(distance, vectorToSource);
-        const targetGain = 0.55 + clarityScore * 0.6; // 0.55 → 1.15
-        const targetLowpass = 3200 + clarityScore * 4200; // 3.2kHz → ~7.4kHz
+        const targetGain = 0.48 + clarityScore * 0.72; // 0.48 → 1.20
+        const targetLowpass = 3600 + clarityScore * 4600; // 3.6kHz → ~8.2kHz
         nodes.proximityGain.gain.setTargetAtTime(targetGain, this.audioContext.currentTime, 0.08);
         nodes.dynamicLowpass.frequency.setTargetAtTime(targetLowpass, this.audioContext.currentTime, 0.12);
         if (Math.random() < 0.005) {
@@ -526,8 +530,8 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         return this.clamp(0.2 + proximityWeight * 0.6 + focusWeight * 0.2, 0, 1);
     }
     calculateProximityWeight(distance) {
-        const closeRange = 1.2;
-        const fadeRange = 12;
+        const closeRange = 0.85;
+        const fadeRange = 18;
         if (distance <= closeRange) {
             return 1;
         }
@@ -598,7 +602,21 @@ class SpatialAudioManager extends EventManager_1.EventManager {
     this.learnRate = this._sanitize(cfg.learnRate, 0.001, 0.3, 0.08);
     this.voiceBoost = this._sanitize(cfg.voiceBoost, 0, 1, 0.6);
     this.voiceSensitivity = this._sanitize(cfg.voiceSensitivity, 0.05, 0.9, 0.35);
-    this.historySize = 512;
+    this.voiceEnhancement = cfg.voiceEnhancement === true;
+    this.silenceFloor = this._sanitize(cfg.silenceFloor, 0.0002, 0.02, 0.0012);
+    this.silenceHoldSamples = Math.max(
+      8,
+      Math.round(
+        sampleRate * this._sanitize(cfg.silenceHoldMs, 40, 1200, 260) / 1000
+      )
+    );
+    this.silenceReleaseSamples = Math.max(
+      8,
+      Math.round(
+        sampleRate * this._sanitize(cfg.silenceReleaseMs, 30, 800, 140) / 1000
+      )
+    );
+    this.historySize = this.voiceEnhancement ? 512 : 0;
     this.channelState = [];
     this.hfAlpha = Math.exp(-2 * Math.PI * 3200 / sampleRate);
   }
@@ -618,17 +636,24 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         gain: 1,
         quietSamples: 0,
         lpState: 0,
-        history: new Float32Array(this.historySize),
+        history: this.voiceEnhancement ? new Float32Array(this.historySize) : null,
         historyIndex: 0,
         historyFilled: 0,
-        tempBuffer: new Float32Array(this.historySize),
+        tempBuffer: this.voiceEnhancement ? new Float32Array(this.historySize) : null,
         voiceConfidence: 0,
+        silenceSamples: 0,
+        silenceReleaseCounter: 0,
+        isSilenced: false,
+        muteGain: 1,
       };
     }
     return this.channelState[index];
   }
 
   _pushHistory(state, sample) {
+    if (!this.voiceEnhancement || !state.history) {
+      return;
+    }
     state.history[state.historyIndex] = sample;
     state.historyIndex = (state.historyIndex + 1) % state.history.length;
     if (state.historyFilled < state.history.length) {
@@ -637,6 +662,11 @@ class SpatialAudioManager extends EventManager_1.EventManager {
   }
 
   _updateVoiceConfidence(state) {
+    if (!this.voiceEnhancement || !state.history || !state.tempBuffer) {
+      state.voiceConfidence += (0 - state.voiceConfidence) * 0.2;
+      return state.voiceConfidence;
+    }
+
     if (state.historyFilled < state.history.length * 0.6) {
       state.voiceConfidence += (0 - state.voiceConfidence) * 0.15;
       return state.voiceConfidence;
@@ -700,7 +730,9 @@ class SpatialAudioManager extends EventManager_1.EventManager {
       }
 
       const state = this._ensureState(channel);
-      const speechPresence = this.voiceBoost * state.voiceConfidence;
+      const speechPresence = this.voiceEnhancement
+        ? this.voiceBoost * state.voiceConfidence
+        : 0;
 
       for (let i = 0; i < inChannel.length; i++) {
         const sample = inChannel[i];
@@ -716,6 +748,33 @@ class SpatialAudioManager extends EventManager_1.EventManager {
           state.quietSamples = 0;
         }
 
+        if (state.envelope < this.silenceFloor && speechPresence < 0.2) {
+          state.silenceSamples++;
+        } else {
+          state.silenceSamples = Math.max(0, state.silenceSamples - 2);
+        }
+
+        if (!state.isSilenced && state.silenceSamples > this.silenceHoldSamples) {
+          state.isSilenced = true;
+          state.silenceReleaseCounter = 0;
+        }
+
+        if (state.isSilenced) {
+          if (
+            state.envelope > this.silenceFloor * 1.8 ||
+            speechPresence > 0.35
+          ) {
+            state.silenceReleaseCounter++;
+            if (state.silenceReleaseCounter > this.silenceReleaseSamples) {
+              state.isSilenced = false;
+              state.silenceSamples = 0;
+              state.silenceReleaseCounter = 0;
+            }
+          } else {
+            state.silenceReleaseCounter = 0;
+          }
+        }
+
         const ratio = state.noise / Math.max(state.envelope, 1e-6);
         let gainTarget = 1 - Math.min(0.98, Math.pow(ratio, this.expansionRatio));
         gainTarget = Math.max(0, Math.min(1, gainTarget));
@@ -724,12 +783,15 @@ class SpatialAudioManager extends EventManager_1.EventManager {
           gainTarget *= 1 - this.maxReduction;
         }
 
-        const reductionFloor = 1 - this.maxReduction * (1 - Math.min(1, speechPresence * 0.85));
+        const reductionFloor = this.voiceEnhancement
+          ? 1 - this.maxReduction * (1 - Math.min(1, speechPresence * 0.85))
+          : 1 - this.maxReduction;
         if (gainTarget < reductionFloor) {
           gainTarget = reductionFloor;
         }
 
-        const dynamicRelease = this.release * (speechPresence > 0.1 ? 0.6 : 1);
+        const dynamicRelease = this.release *
+          (this.voiceEnhancement && speechPresence > 0.1 ? 0.6 : 1);
         state.gain += (gainTarget - state.gain) * dynamicRelease;
         let processed = sample * state.gain;
 
@@ -742,6 +804,10 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         const hissGain = 1 - hissRatio * (this.hissCut * (1 - 0.4 * speechPresence));
         processed = state.lpState + highComponent * hissGain;
 
+        const muteTarget = state.isSilenced ? 0 : 1;
+        state.muteGain += (muteTarget - state.muteGain) * 0.35;
+        processed *= state.muteGain;
+
         outChannel[i] = processed;
       }
 
@@ -779,17 +845,21 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
         };
         const denoiserDefaults = {
             enabled: true,
-            threshold: 0.009,
-            noiseFloor: 0.0025,
-            release: 0.24,
-            attack: 0.25,
-            holdMs: 150,
-            maxReduction: 0.92,
-            hissCut: 0.62,
-            expansionRatio: 2.35,
-            learnRate: 0.06,
+            threshold: 0.0082,
+            noiseFloor: 0.0022,
+            release: 0.28,
+            attack: 0.32,
+            holdMs: 190,
+            maxReduction: 0.85,
+            hissCut: 0.5,
+            expansionRatio: 2.1,
+            learnRate: 0.05,
             voiceBoost: 0.6,
             voiceSensitivity: 0.35,
+            voiceEnhancement: false,
+            silenceFloor: 0.0006,
+            silenceHoldMs: 420,
+            silenceReleaseMs: 260,
         };
         return {
             distance: {
@@ -811,6 +881,10 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
                 learnRate: options?.denoiser?.learnRate ?? denoiserDefaults.learnRate,
                 voiceBoost: options?.denoiser?.voiceBoost ?? denoiserDefaults.voiceBoost,
                 voiceSensitivity: options?.denoiser?.voiceSensitivity ?? denoiserDefaults.voiceSensitivity,
+                voiceEnhancement: options?.denoiser?.voiceEnhancement ?? denoiserDefaults.voiceEnhancement,
+                silenceFloor: options?.denoiser?.silenceFloor ?? denoiserDefaults.silenceFloor,
+                silenceHoldMs: options?.denoiser?.silenceHoldMs ?? denoiserDefaults.silenceHoldMs,
+                silenceReleaseMs: options?.denoiser?.silenceReleaseMs ?? denoiserDefaults.silenceReleaseMs,
             },
         };
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@newgameplusinc/odyssey-audio-video-sdk-dev",
-  "version": "1.0.20",
+  "version": "1.0.22",
   "description": "Odyssey Spatial Audio & Video SDK using MediaSoup for real-time communication",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",