@newgameplusinc/odyssey-audio-video-sdk-dev 1.0.19 → 1.0.21

package/README.md

@@ -79,7 +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,
+			voiceBoost: 0.7,
+			voiceSensitivity: 0.3,
+			voiceEnhancement: true,
+			silenceFloor: 0.001,
+			silenceHoldMs: 220,
+			silenceReleaseMs: 160,
+		},
+	});
+	```
+	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. Tweak these knobs if you need even more “AirPods Pro” style isolation.
 
 #### 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

@@ -17,8 +17,14 @@ type DenoiserOptions = {
     hissCut?: number;
     expansionRatio?: number;
     learnRate?: number;
+    voiceBoost?: number;
+    voiceSensitivity?: number;
+    voiceEnhancement?: boolean;
+    silenceFloor?: number;
+    silenceHoldMs?: number;
+    silenceReleaseMs?: number;
 };
-type SpatialAudioOptions = {
+export type SpatialAudioOptions = {
     distance?: SpatialAudioDistanceConfig;
     denoiser?: DenoiserOptions;
 };

package/dist/SpatialAudioManager.js

@@ -88,6 +88,12 @@ class SpatialAudioManager extends EventManager_1.EventManager {
                         hissCut: this.options.denoiser?.hissCut,
                         expansionRatio: this.options.denoiser?.expansionRatio,
                         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,
                     },
                 });
             }
@@ -594,6 +600,23 @@ class SpatialAudioManager extends EventManager_1.EventManager {
     this.hissCut = this._sanitize(cfg.hissCut, 0, 1, 0.45);
     this.expansionRatio = this._sanitize(cfg.expansionRatio, 1.1, 4, 1.8);
     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.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);
   }
@@ -613,11 +636,78 @@ class SpatialAudioManager extends EventManager_1.EventManager {
         gain: 1,
         quietSamples: 0,
         lpState: 0,
+        history: this.voiceEnhancement ? new Float32Array(this.historySize) : null,
+        historyIndex: 0,
+        historyFilled: 0,
+        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) {
+      state.historyFilled++;
+    }
+  }
+
+  _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;
+    }
+
+    const len = state.history.length;
+    let writeIndex = state.historyIndex;
+    for (let i = 0; i < len; i++) {
+      state.tempBuffer[i] = state.history[writeIndex];
+      writeIndex = (writeIndex + 1) % len;
+    }
+
+    const minLag = 30;
+    const maxLag = 240;
+    let best = 0;
+    for (let lag = minLag; lag <= maxLag; lag += 2) {
+      let sum = 0;
+      let energyA = 0;
+      let energyB = 0;
+      for (let i = lag; i < len; i++) {
+        const a = state.tempBuffer[i];
+        const b = state.tempBuffer[i - lag];
+        sum += a * b;
+        energyA += a * a;
+        energyB += b * b;
+      }
+      const denom = Math.sqrt(energyA * energyB) + 1e-8;
+      const corr = Math.abs(sum) / denom;
+      if (corr > best) {
+        best = corr;
+      }
+    }
+
+    const normalized = Math.max(
+      0,
+      Math.min(1, (best - this.voiceSensitivity) / (1 - this.voiceSensitivity))
+    );
+    state.voiceConfidence += (normalized - state.voiceConfidence) * 0.2;
+    return state.voiceConfidence;
+  }
+
   process(inputs, outputs) {
     const input = inputs[0];
     const output = outputs[0];
@@ -640,9 +730,13 @@ class SpatialAudioManager extends EventManager_1.EventManager {
       }
 
       const state = this._ensureState(channel);
+      const speechPresence = this.voiceEnhancement
+        ? this.voiceBoost * state.voiceConfidence
+        : 0;
 
       for (let i = 0; i < inChannel.length; i++) {
         const sample = inChannel[i];
+        this._pushHistory(state, sample);
         const magnitude = Math.abs(sample);
 
         state.envelope += (magnitude - state.envelope) * this.attack;
@@ -654,6 +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));
@@ -662,7 +783,16 @@ class SpatialAudioManager extends EventManager_1.EventManager {
           gainTarget *= 1 - this.maxReduction;
         }
 
-        state.gain += (gainTarget - state.gain) * this.release;
+        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 *
+          (this.voiceEnhancement && speechPresence > 0.1 ? 0.6 : 1);
+        state.gain += (gainTarget - state.gain) * dynamicRelease;
         let processed = sample * state.gain;
 
         state.lpState = this.hfAlpha * state.lpState + (1 - this.hfAlpha) * processed;
@@ -671,11 +801,17 @@ class SpatialAudioManager extends EventManager_1.EventManager {
           1,
           Math.abs(highComponent) / (Math.abs(state.lpState) + 1e-5)
         );
-        const hissGain = 1 - hissRatio * this.hissCut;
+        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;
       }
+
+      this._updateVoiceConfidence(state);
     }
 
     return true;
@@ -709,15 +845,21 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
         };
         const denoiserDefaults = {
             enabled: true,
-            threshold: 0.012,
-            noiseFloor: 0.004,
-            release: 0.18,
-            attack: 0.35,
-            holdMs: 110,
-            maxReduction: 0.85,
-            hissCut: 0.45,
-            expansionRatio: 1.8,
-            learnRate: 0.08,
+            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,
+            voiceBoost: 0.6,
+            voiceSensitivity: 0.35,
+            voiceEnhancement: false,
+            silenceFloor: 0.0012,
+            silenceHoldMs: 260,
+            silenceReleaseMs: 140,
         };
         return {
             distance: {
@@ -737,6 +879,12 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
                 hissCut: options?.denoiser?.hissCut ?? denoiserDefaults.hissCut,
                 expansionRatio: options?.denoiser?.expansionRatio ?? denoiserDefaults.expansionRatio,
                 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/dist/index.d.ts

@@ -1,4 +1,5 @@
 import { EventManager } from "./EventManager";
+import { SpatialAudioOptions } from "./SpatialAudioManager";
 import { Direction, MediaState, OdysseyEvent, Participant, Position, RoomJoinedData } from "./types";
 export declare class OdysseySpatialComms extends EventManager {
     private socket;
@@ -9,7 +10,7 @@ export declare class OdysseySpatialComms extends EventManager {
     private localParticipant;
     private mediasoupManager;
     private spatialAudioManager;
-    constructor(serverUrl: string);
+    constructor(serverUrl: string, spatialOptions?: SpatialAudioOptions);
     on(event: OdysseyEvent, listener: (...args: any[]) => void): this;
     emit(event: OdysseyEvent, ...args: any[]): boolean;
     joinRoom(data: {

package/dist/index.js

@@ -6,7 +6,7 @@ const EventManager_1 = require("./EventManager");
 const MediasoupManager_1 = require("./MediasoupManager");
 const SpatialAudioManager_1 = require("./SpatialAudioManager");
 class OdysseySpatialComms extends EventManager_1.EventManager {
-    constructor(serverUrl) {
+    constructor(serverUrl, spatialOptions) {
         super(); // Initialize the EventEmitter base class
         this.room = null;
         this.localParticipant = null;
@@ -14,7 +14,7 @@ class OdysseySpatialComms extends EventManager_1.EventManager {
             transports: ["websocket"],
         });
         this.mediasoupManager = new MediasoupManager_1.MediasoupManager(this.socket);
-        this.spatialAudioManager = new SpatialAudioManager_1.SpatialAudioManager();
+        this.spatialAudioManager = new SpatialAudioManager_1.SpatialAudioManager(spatialOptions);
         // Set max listeners to prevent warning
         this.setMaxListeners(50);
         this.listenForEvents();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@newgameplusinc/odyssey-audio-video-sdk-dev",
-  "version": "1.0.19",
+  "version": "1.0.21",
   "description": "Odyssey Spatial Audio & Video SDK using MediaSoup for real-time communication",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",