@newgameplusinc/odyssey-audio-video-sdk-dev 1.0.18 → 1.0.20

package/README.md

@@ -79,7 +79,20 @@ 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** – optional AudioWorklet denoiser plus high/low-pass filters trim rumble & hiss before HRTF processing.
+- **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.
+	```ts
+	const sdk = new OdysseySpatialComms(serverUrl, {
+		denoiser: {
+			threshold: 0.0085,
+			maxReduction: 0.94,
+			hissCut: 0.7,
+			holdMs: 180,
+			voiceBoost: 0.7,
+			voiceSensitivity: 0.3,
+		},
+	});
+	```
+	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

@@ -11,8 +11,16 @@ type DenoiserOptions = {
     threshold?: number;
     noiseFloor?: number;
     release?: number;
+    attack?: number;
+    holdMs?: number;
+    maxReduction?: number;
+    hissCut?: number;
+    expansionRatio?: number;
+    learnRate?: number;
+    voiceBoost?: number;
+    voiceSensitivity?: number;
 };
-type SpatialAudioOptions = {
+export type SpatialAudioOptions = {
     distance?: SpatialAudioDistanceConfig;
     denoiser?: DenoiserOptions;
 };
@@ -25,7 +33,6 @@ export declare class SpatialAudioManager extends EventManager {
     private options;
     private denoiseWorkletReady;
     private denoiseWorkletUrl?;
-    private denoiserWasmBytes?;
     private listenerPosition;
     private listenerInitialized;
     private listenerDirection;

package/dist/SpatialAudioManager.js

@@ -82,9 +82,14 @@ class SpatialAudioManager extends EventManager_1.EventManager {
                         threshold: this.options.denoiser?.threshold,
                         noiseFloor: this.options.denoiser?.noiseFloor,
                         release: this.options.denoiser?.release,
-                        wasmBytes: this.denoiserWasmBytes
-                            ? this.denoiserWasmBytes.slice(0)
-                            : null,
+                        attack: this.options.denoiser?.attack,
+                        holdMs: this.options.denoiser?.holdMs,
+                        maxReduction: this.options.denoiser?.maxReduction,
+                        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,
                     },
                 });
             }
@@ -577,48 +582,170 @@ class SpatialAudioManager extends EventManager_1.EventManager {
     super();
     const cfg = (options && options.processorOptions) || {};
     this.enabled = cfg.enabled !== false;
-    this.threshold = typeof cfg.threshold === 'number' ? cfg.threshold : 0.012;
-    this.noiseFloor = typeof cfg.noiseFloor === 'number' ? cfg.noiseFloor : 0.004;
-    this.release = typeof cfg.release === 'number' ? cfg.release : 0.18;
-    this.smoothedLevel = this.noiseFloor;
+    this.threshold = this._sanitize(cfg.threshold, 0.003, 0.05, 0.012);
+    this.noiseFloor = this._sanitize(cfg.noiseFloor, 0.0005, 0.05, 0.004);
+    this.attack = this._sanitize(cfg.attack, 0.01, 0.9, 0.35);
+    this.release = this._sanitize(cfg.release, 0.01, 0.9, 0.18);
+    this.holdSamples = Math.max(
+      8,
+      Math.round(
+        sampleRate * this._sanitize(cfg.holdMs, 10, 400, 110) / 1000
+      )
+    );
+    this.maxReduction = this._sanitize(cfg.maxReduction, 0.1, 0.95, 0.85);
+    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.historySize = 512;
+    this.channelState = [];
+    this.hfAlpha = Math.exp(-2 * Math.PI * 3200 / sampleRate);
+  }
+
+  _sanitize(value, min, max, fallback) {
+    if (typeof value !== 'number' || !isFinite(value)) {
+      return fallback;
+    }
+    return Math.min(max, Math.max(min, value));
+  }
+
+  _ensureState(index) {
+    if (!this.channelState[index]) {
+      this.channelState[index] = {
+        envelope: this.noiseFloor,
+        noise: this.noiseFloor,
+        gain: 1,
+        quietSamples: 0,
+        lpState: 0,
+        history: new Float32Array(this.historySize),
+        historyIndex: 0,
+        historyFilled: 0,
+        tempBuffer: new Float32Array(this.historySize),
+        voiceConfidence: 0,
+      };
+    }
+    return this.channelState[index];
+  }
+
+  _pushHistory(state, sample) {
+    state.history[state.historyIndex] = sample;
+    state.historyIndex = (state.historyIndex + 1) % state.history.length;
+    if (state.historyFilled < state.history.length) {
+      state.historyFilled++;
+    }
+  }
+
+  _updateVoiceConfidence(state) {
+    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];
-    if (!input || input.length === 0 || !output || output.length === 0) {
+    if (!input || !output) {
       return true;
     }
 
-    for (let channel = 0; channel < input.length; channel++) {
+    for (let channel = 0; channel < output.length; channel++) {
       const inChannel = input[channel];
       const outChannel = output[channel];
       if (!inChannel || !outChannel) {
         continue;
       }
 
-      let sum = 0;
+      if (!this.enabled) {
+        for (let i = 0; i < inChannel.length; i++) {
+          outChannel[i] = inChannel[i];
+        }
+        continue;
+      }
+
+      const state = this._ensureState(channel);
+      const speechPresence = this.voiceBoost * state.voiceConfidence;
+
       for (let i = 0; i < inChannel.length; i++) {
         const sample = inChannel[i];
-        sum += sample * sample;
-      }
+        this._pushHistory(state, sample);
+        const magnitude = Math.abs(sample);
 
-      const rms = Math.sqrt(sum / inChannel.length);
-      this.smoothedLevel += (rms - this.smoothedLevel) * this.release;
-      const dynamicThreshold = Math.max(
-        this.noiseFloor,
-        this.threshold * 0.6 + this.smoothedLevel * 0.4
-      );
+        state.envelope += (magnitude - state.envelope) * this.attack;
 
-      if (!this.enabled || rms >= dynamicThreshold) {
-        for (let i = 0; i < inChannel.length; i++) {
-          outChannel[i] = inChannel[i];
+        if (state.envelope < this.threshold) {
+          state.noise += (state.envelope - state.noise) * this.learnRate;
+          state.quietSamples++;
+        } else {
+          state.quietSamples = 0;
         }
-      } else {
-        for (let i = 0; i < inChannel.length; i++) {
-          outChannel[i] = 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));
+
+        if (state.quietSamples > this.holdSamples) {
+          gainTarget *= 1 - this.maxReduction;
+        }
+
+        const reductionFloor = 1 - this.maxReduction * (1 - Math.min(1, speechPresence * 0.85));
+        if (gainTarget < reductionFloor) {
+          gainTarget = reductionFloor;
         }
+
+        const dynamicRelease = this.release * (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;
+        const highComponent = processed - state.lpState;
+        const hissRatio = Math.min(
+          1,
+          Math.abs(highComponent) / (Math.abs(state.lpState) + 1e-5)
+        );
+        const hissGain = 1 - hissRatio * (this.hissCut * (1 - 0.4 * speechPresence));
+        processed = state.lpState + highComponent * hissGain;
+
+        outChannel[i] = processed;
       }
+
+      this._updateVoiceConfidence(state);
     }
 
     return true;
@@ -652,9 +779,17 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
         };
         const denoiserDefaults = {
             enabled: true,
-            threshold: 0.012,
-            noiseFloor: 0.004,
-            release: 0.18,
+            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,
         };
         return {
             distance: {
@@ -668,6 +803,14 @@ registerProcessor('odyssey-denoise', OdysseyDenoiseProcessor);
                 threshold: options?.denoiser?.threshold ?? denoiserDefaults.threshold,
                 noiseFloor: options?.denoiser?.noiseFloor ?? denoiserDefaults.noiseFloor,
                 release: options?.denoiser?.release ?? denoiserDefaults.release,
+                attack: options?.denoiser?.attack ?? denoiserDefaults.attack,
+                holdMs: options?.denoiser?.holdMs ?? denoiserDefaults.holdMs,
+                maxReduction: options?.denoiser?.maxReduction ?? denoiserDefaults.maxReduction,
+                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,
             },
         };
     }

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": "1.0.18",
+  "version": "1.0.20",
   "description": "Odyssey Spatial Audio & Video SDK using MediaSoup for real-time communication",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",