@tensamin/audio 0.1.0 → 0.1.2

package/dist/pipeline/audio-pipeline.js

@@ -0,0 +1,335 @@
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/pipeline/audio-pipeline.ts
+var audio_pipeline_exports = {};
+__export(audio_pipeline_exports, {
+  createAudioPipeline: () => createAudioPipeline
+});
+module.exports = __toCommonJS(audio_pipeline_exports);
+var import_mitt = __toESM(require("mitt"));
+
+// src/context/audio-context.ts
+var sharedContext = null;
+var activePipelines = 0;
+function getAudioContext(options) {
+  if (typeof window === "undefined" || typeof AudioContext === "undefined") {
+    throw new Error(
+      "AudioContext is not supported in this environment (browser only)."
+    );
+  }
+  if (!sharedContext || sharedContext.state === "closed") {
+    sharedContext = new AudioContext(options);
+  }
+  return sharedContext;
+}
+function registerPipeline() {
+  activePipelines++;
+}
+function unregisterPipeline() {
+  activePipelines = Math.max(0, activePipelines - 1);
+}
+
+// src/noise-suppression/rnnoise-node.ts
+var RNNoisePlugin = class {
+  name = "rnnoise-ns";
+  wasmBuffer = null;
+  async createNode(context, config) {
+    const { loadRnnoise, RnnoiseWorkletNode } = await import("@sapphi-red/web-noise-suppressor");
+    if (!config?.enabled) {
+      const pass = context.createGain();
+      return pass;
+    }
+    if (!config?.wasmUrl || !config?.simdUrl || !config?.workletUrl) {
+      throw new Error(
+        "RNNoisePlugin requires 'wasmUrl', 'simdUrl', and 'workletUrl' to be configured. Please download the assets and provide the URLs."
+      );
+    }
+    if (!this.wasmBuffer) {
+      this.wasmBuffer = await loadRnnoise({
+        url: config.wasmUrl,
+        simdUrl: config.simdUrl
+      });
+    }
+    const workletUrl = config.workletUrl;
+    try {
+      await context.audioWorklet.addModule(workletUrl);
+    } catch (e) {
+      console.warn("Failed to add RNNoise worklet module:", e);
+    }
+    const node = new RnnoiseWorkletNode(context, {
+      wasmBinary: this.wasmBuffer,
+      maxChannels: 1
+      // Mono for now
+    });
+    return node;
+  }
+};
+
+// src/vad/vad-node.ts
+var energyVadWorkletCode = `
+class EnergyVadProcessor extends AudioWorkletProcessor {
+  constructor() {
+    super();
+    this.smoothing = 0.95;
+    this.energy = 0;
+    this.noiseFloor = 0.001;
+  }
+
+  process(inputs, outputs, parameters) {
+    const input = inputs[0];
+    if (!input || !input.length) return true;
+    const channel = input[0];
+    
+    // Calculate RMS
+    let sum = 0;
+    for (let i = 0; i < channel.length; i++) {
+      sum += channel[i] * channel[i];
+    }
+    const rms = Math.sqrt(sum / channel.length);
+
+    // Simple adaptive noise floor (very basic)
+    if (rms < this.noiseFloor) {
+      this.noiseFloor = this.noiseFloor * 0.99 + rms * 0.01;
+    } else {
+      this.noiseFloor = this.noiseFloor * 0.999 + rms * 0.001;
+    }
+
+    // Calculate "probability" based on SNR
+    // This is a heuristic mapping from energy to 0-1
+    const snr = rms / (this.noiseFloor + 1e-6);
+    const probability = Math.min(1, Math.max(0, (snr - 1.5) / 10)); // Arbitrary scaling
+
+    this.port.postMessage({ probability });
+
+    return true;
+  }
+}
+registerProcessor('energy-vad-processor', EnergyVadProcessor);
+`;
+var EnergyVADPlugin = class {
+  name = "energy-vad";
+  async createNode(context, config, onDecision) {
+    const blob = new Blob([energyVadWorkletCode], {
+      type: "application/javascript"
+    });
+    const url = URL.createObjectURL(blob);
+    try {
+      await context.audioWorklet.addModule(url);
+    } catch (e) {
+      console.warn("Failed to add Energy VAD worklet:", e);
+      throw e;
+    } finally {
+      URL.revokeObjectURL(url);
+    }
+    const node = new AudioWorkletNode(context, "energy-vad-processor");
+    node.port.onmessage = (event) => {
+      const { probability } = event.data;
+      onDecision(probability);
+    };
+    return node;
+  }
+};
+
+// src/extensibility/plugins.ts
+var nsPlugins = /* @__PURE__ */ new Map();
+var vadPlugins = /* @__PURE__ */ new Map();
+var defaultNs = new RNNoisePlugin();
+nsPlugins.set(defaultNs.name, defaultNs);
+var defaultVad = new EnergyVADPlugin();
+vadPlugins.set(defaultVad.name, defaultVad);
+function getNoiseSuppressionPlugin(name) {
+  if (!name) return defaultNs;
+  const plugin = nsPlugins.get(name);
+  if (!plugin) {
+    console.warn(
+      `Noise suppression plugin '${name}' not found, falling back to default.`
+    );
+    return defaultNs;
+  }
+  return plugin;
+}
+function getVADPlugin(name) {
+  if (!name) return defaultVad;
+  const plugin = vadPlugins.get(name);
+  if (!plugin) {
+    console.warn(`VAD plugin '${name}' not found, falling back to default.`);
+    return defaultVad;
+  }
+  return plugin;
+}
+
+// src/vad/vad-state.ts
+var VADStateMachine = class {
+  config;
+  currentState = "silent";
+  lastSpeechTime = 0;
+  speechStartTime = 0;
+  frameDurationMs = 20;
+  // Assumed frame duration, updated by calls
+  constructor(config) {
+    this.config = {
+      enabled: config?.enabled ?? true,
+      pluginName: config?.pluginName ?? "energy-vad",
+      startThreshold: config?.startThreshold ?? 0.5,
+      stopThreshold: config?.stopThreshold ?? 0.4,
+      hangoverMs: config?.hangoverMs ?? 300,
+      preRollMs: config?.preRollMs ?? 200
+    };
+  }
+  updateConfig(config) {
+    this.config = { ...this.config, ...config };
+  }
+  processFrame(probability, timestamp) {
+    const { startThreshold, stopThreshold, hangoverMs } = this.config;
+    let newState = this.currentState;
+    if (this.currentState === "silent" || this.currentState === "speech_ending") {
+      if (probability >= startThreshold) {
+        newState = "speech_starting";
+        this.speechStartTime = timestamp;
+        this.lastSpeechTime = timestamp;
+      } else {
+        newState = "silent";
+      }
+    } else if (this.currentState === "speech_starting" || this.currentState === "speaking") {
+      if (probability >= stopThreshold) {
+        newState = "speaking";
+        this.lastSpeechTime = timestamp;
+      } else {
+        const timeSinceSpeech = timestamp - this.lastSpeechTime;
+        if (timeSinceSpeech < hangoverMs) {
+          newState = "speaking";
+        } else {
+          newState = "speech_ending";
+        }
+      }
+    }
+    if (newState === "speech_starting") newState = "speaking";
+    if (newState === "speech_ending") newState = "silent";
+    this.currentState = newState;
+    return {
+      isSpeaking: newState === "speaking",
+      probability,
+      state: newState
+    };
+  }
+};
+
+// src/pipeline/audio-pipeline.ts
+async function createAudioPipeline(sourceTrack, config = {}) {
+  const context = getAudioContext();
+  registerPipeline();
+  const fullConfig = {
+    noiseSuppression: { enabled: true, ...config.noiseSuppression },
+    vad: { enabled: true, ...config.vad },
+    output: {
+      speechGain: 1,
+      silenceGain: 0,
+      gainRampTime: 0.02,
+      ...config.output
+    },
+    livekit: { manageTrackMute: false, ...config.livekit }
+  };
+  const sourceStream = new MediaStream([sourceTrack]);
+  const sourceNode = context.createMediaStreamSource(sourceStream);
+  const nsPlugin = getNoiseSuppressionPlugin(
+    fullConfig.noiseSuppression?.pluginName
+  );
+  const nsNode = await nsPlugin.createNode(
+    context,
+    fullConfig.noiseSuppression
+  );
+  const vadPlugin = getVADPlugin(fullConfig.vad?.pluginName);
+  const vadStateMachine = new VADStateMachine(fullConfig.vad);
+  const emitter = (0, import_mitt.default)();
+  const vadNode = await vadPlugin.createNode(
+    context,
+    fullConfig.vad,
+    (prob) => {
+      const timestamp = context.currentTime * 1e3;
+      const newState = vadStateMachine.processFrame(prob, timestamp);
+      if (newState.state !== lastVadState.state || Math.abs(newState.probability - lastVadState.probability) > 0.1) {
+        emitter.emit("vadChange", newState);
+        lastVadState = newState;
+        updateGain(newState);
+      }
+    }
+  );
+  let lastVadState = {
+    isSpeaking: false,
+    probability: 0,
+    state: "silent"
+  };
+  const splitter = context.createGain();
+  sourceNode.connect(nsNode);
+  nsNode.connect(splitter);
+  splitter.connect(vadNode);
+  const delayNode = context.createDelay(1);
+  const preRollSeconds = (fullConfig.vad?.preRollMs ?? 200) / 1e3;
+  delayNode.delayTime.value = preRollSeconds;
+  const gainNode = context.createGain();
+  gainNode.gain.value = fullConfig.output?.silenceGain ?? 0;
+  const destination = context.createMediaStreamDestination();
+  splitter.connect(delayNode);
+  delayNode.connect(gainNode);
+  gainNode.connect(destination);
+  function updateGain(state) {
+    const { speechGain, silenceGain, gainRampTime } = fullConfig.output;
+    const targetGain = state.isSpeaking ? speechGain ?? 1 : silenceGain ?? 0;
+    const now = context.currentTime;
+    gainNode.gain.setTargetAtTime(targetGain, now, gainRampTime ?? 0.02);
+  }
+  function dispose() {
+    sourceNode.disconnect();
+    nsNode.disconnect();
+    splitter.disconnect();
+    vadNode.disconnect();
+    delayNode.disconnect();
+    gainNode.disconnect();
+    destination.stream.getTracks().forEach((t) => t.stop());
+    unregisterPipeline();
+  }
+  return {
+    processedTrack: destination.stream.getAudioTracks()[0],
+    events: emitter,
+    get state() {
+      return lastVadState;
+    },
+    setConfig: (newConfig) => {
+      if (newConfig.vad) {
+        vadStateMachine.updateConfig(newConfig.vad);
+      }
+    },
+    dispose
+  };
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  createAudioPipeline
+});

package/dist/pipeline/audio-pipeline.mjs

@@ -0,0 +1,11 @@
+import {
+  createAudioPipeline
+} from "../chunk-QU7E5HBA.mjs";
+import "../chunk-JJASCVEW.mjs";
+import "../chunk-OZ7KMC4S.mjs";
+import "../chunk-FS635GMR.mjs";
+import "../chunk-SDTOKWM2.mjs";
+import "../chunk-UMU2KIB6.mjs";
+export {
+  createAudioPipeline
+};

package/dist/types.d.mts

@@ -0,0 +1,155 @@
+import { Emitter } from 'mitt';
+
+/**
+ * Configuration for the audio processing pipeline.
+ */
+interface AudioProcessingConfig {
+    /**
+     * Noise suppression configuration.
+     */
+    noiseSuppression?: {
+        enabled: boolean;
+        /**
+         * Path or URL to the RNNoise WASM binary.
+         * REQUIRED if enabled.
+         */
+        wasmUrl?: string;
+        /**
+         * Path or URL to the RNNoise SIMD WASM binary.
+         * REQUIRED if enabled.
+         */
+        simdUrl?: string;
+        /**
+         * Path or URL to the RNNoise worklet script.
+         * REQUIRED if enabled.
+         */
+        workletUrl?: string;
+        /**
+         * Plugin name to use. Defaults to 'rnnoise-ns'.
+         */
+        pluginName?: string;
+    };
+    /**
+     * Voice Activity Detection (VAD) configuration.
+     */
+    vad?: {
+        enabled: boolean;
+        /**
+         * Plugin name to use. Defaults to 'rnnoise-vad' or 'energy-vad'.
+         */
+        pluginName?: string;
+        /**
+         * Probability threshold for speech onset (0-1).
+         * Default: 0.5
+         */
+        startThreshold?: number;
+        /**
+         * Probability threshold for speech offset (0-1).
+         * Default: 0.4
+         */
+        stopThreshold?: number;
+        /**
+         * Time in ms to wait after speech stops before considering it silent.
+         * Default: 300ms
+         */
+        hangoverMs?: number;
+        /**
+         * Time in ms of audio to buffer before speech onset to avoid cutting the start.
+         * Default: 200ms
+         */
+        preRollMs?: number;
+    };
+    /**
+     * Output gain and muting configuration.
+     */
+    output?: {
+        /**
+         * Gain to apply when speaking (0-1+). Default: 1.0
+         */
+        speechGain?: number;
+        /**
+         * Gain to apply when silent (0-1). Default: 0.0 (mute)
+         */
+        silenceGain?: number;
+        /**
+         * Time in seconds to ramp gain changes. Default: 0.02
+         */
+        gainRampTime?: number;
+    };
+    /**
+     * LiveKit integration configuration.
+     */
+    livekit?: {
+        /**
+         * Whether to call track.mute()/unmute() on the LocalAudioTrack based on VAD.
+         * This saves bandwidth but has more signaling overhead.
+         * Default: false (uses gain gating only)
+         */
+        manageTrackMute?: boolean;
+    };
+}
+/**
+ * Represents the state of Voice Activity Detection.
+ */
+interface VADState {
+    /**
+     * Whether speech is currently detected (after hysteresis).
+     */
+    isSpeaking: boolean;
+    /**
+     * Raw probability of speech from the VAD model (0-1).
+     */
+    probability: number;
+    /**
+     * Current state enum.
+     */
+    state: "silent" | "speech_starting" | "speaking" | "speech_ending";
+}
+/**
+ * Events emitted by the audio pipeline.
+ */
+type AudioPipelineEvents = {
+    vadChange: VADState;
+    error: Error;
+};
+/**
+ * Handle to a running audio processing pipeline.
+ */
+interface AudioPipelineHandle {
+    /**
+     * The processed MediaStreamTrack.
+     */
+    readonly processedTrack: MediaStreamTrack;
+    /**
+     * Event emitter for VAD state and errors.
+     */
+    readonly events: Emitter<AudioPipelineEvents>;
+    /**
+     * Current VAD state.
+     */
+    readonly state: VADState;
+    /**
+     * Update configuration at runtime.
+     */
+    setConfig(config: Partial<AudioProcessingConfig>): void;
+    /**
+     * Stop processing and release resources.
+     */
+    dispose(): void;
+}
+/**
+ * Interface for a Noise Suppression Plugin.
+ */
+interface NoiseSuppressionPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["noiseSuppression"]): Promise<AudioNode>;
+}
+/**
+ * Interface for a VAD Plugin.
+ */
+interface VADPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["vad"], onDecision: (probability: number) => void): Promise<AudioNode>;
+}
+
+export type { AudioPipelineEvents, AudioPipelineHandle, AudioProcessingConfig, NoiseSuppressionPlugin, VADPlugin, VADState };

package/dist/types.d.ts

@@ -0,0 +1,155 @@
+import { Emitter } from 'mitt';
+
+/**
+ * Configuration for the audio processing pipeline.
+ */
+interface AudioProcessingConfig {
+    /**
+     * Noise suppression configuration.
+     */
+    noiseSuppression?: {
+        enabled: boolean;
+        /**
+         * Path or URL to the RNNoise WASM binary.
+         * REQUIRED if enabled.
+         */
+        wasmUrl?: string;
+        /**
+         * Path or URL to the RNNoise SIMD WASM binary.
+         * REQUIRED if enabled.
+         */
+        simdUrl?: string;
+        /**
+         * Path or URL to the RNNoise worklet script.
+         * REQUIRED if enabled.
+         */
+        workletUrl?: string;
+        /**
+         * Plugin name to use. Defaults to 'rnnoise-ns'.
+         */
+        pluginName?: string;
+    };
+    /**
+     * Voice Activity Detection (VAD) configuration.
+     */
+    vad?: {
+        enabled: boolean;
+        /**
+         * Plugin name to use. Defaults to 'rnnoise-vad' or 'energy-vad'.
+         */
+        pluginName?: string;
+        /**
+         * Probability threshold for speech onset (0-1).
+         * Default: 0.5
+         */
+        startThreshold?: number;
+        /**
+         * Probability threshold for speech offset (0-1).
+         * Default: 0.4
+         */
+        stopThreshold?: number;
+        /**
+         * Time in ms to wait after speech stops before considering it silent.
+         * Default: 300ms
+         */
+        hangoverMs?: number;
+        /**
+         * Time in ms of audio to buffer before speech onset to avoid cutting the start.
+         * Default: 200ms
+         */
+        preRollMs?: number;
+    };
+    /**
+     * Output gain and muting configuration.
+     */
+    output?: {
+        /**
+         * Gain to apply when speaking (0-1+). Default: 1.0
+         */
+        speechGain?: number;
+        /**
+         * Gain to apply when silent (0-1). Default: 0.0 (mute)
+         */
+        silenceGain?: number;
+        /**
+         * Time in seconds to ramp gain changes. Default: 0.02
+         */
+        gainRampTime?: number;
+    };
+    /**
+     * LiveKit integration configuration.
+     */
+    livekit?: {
+        /**
+         * Whether to call track.mute()/unmute() on the LocalAudioTrack based on VAD.
+         * This saves bandwidth but has more signaling overhead.
+         * Default: false (uses gain gating only)
+         */
+        manageTrackMute?: boolean;
+    };
+}
+/**
+ * Represents the state of Voice Activity Detection.
+ */
+interface VADState {
+    /**
+     * Whether speech is currently detected (after hysteresis).
+     */
+    isSpeaking: boolean;
+    /**
+     * Raw probability of speech from the VAD model (0-1).
+     */
+    probability: number;
+    /**
+     * Current state enum.
+     */
+    state: "silent" | "speech_starting" | "speaking" | "speech_ending";
+}
+/**
+ * Events emitted by the audio pipeline.
+ */
+type AudioPipelineEvents = {
+    vadChange: VADState;
+    error: Error;
+};
+/**
+ * Handle to a running audio processing pipeline.
+ */
+interface AudioPipelineHandle {
+    /**
+     * The processed MediaStreamTrack.
+     */
+    readonly processedTrack: MediaStreamTrack;
+    /**
+     * Event emitter for VAD state and errors.
+     */
+    readonly events: Emitter<AudioPipelineEvents>;
+    /**
+     * Current VAD state.
+     */
+    readonly state: VADState;
+    /**
+     * Update configuration at runtime.
+     */
+    setConfig(config: Partial<AudioProcessingConfig>): void;
+    /**
+     * Stop processing and release resources.
+     */
+    dispose(): void;
+}
+/**
+ * Interface for a Noise Suppression Plugin.
+ */
+interface NoiseSuppressionPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["noiseSuppression"]): Promise<AudioNode>;
+}
+/**
+ * Interface for a VAD Plugin.
+ */
+interface VADPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["vad"], onDecision: (probability: number) => void): Promise<AudioNode>;
+}
+
+export type { AudioPipelineEvents, AudioPipelineHandle, AudioProcessingConfig, NoiseSuppressionPlugin, VADPlugin, VADState };

package/dist/types.js

@@ -0,0 +1,18 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/types.ts
+var types_exports = {};
+module.exports = __toCommonJS(types_exports);

package/dist/types.mjs

@@ -0,0 +1 @@
+import "./chunk-WBQAMGXK.mjs";

package/dist/vad/vad-node.d.mts

@@ -0,0 +1,9 @@
+import { VADPlugin, AudioProcessingConfig } from '../types.mjs';
+import 'mitt';
+
+declare class EnergyVADPlugin implements VADPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["vad"], onDecision: (probability: number) => void): Promise<AudioNode>;
+}
+
+export { EnergyVADPlugin };

package/dist/vad/vad-node.d.ts

@@ -0,0 +1,9 @@
+import { VADPlugin, AudioProcessingConfig } from '../types.js';
+import 'mitt';
+
+declare class EnergyVADPlugin implements VADPlugin {
+    name: string;
+    createNode(context: AudioContext, config: AudioProcessingConfig["vad"], onDecision: (probability: number) => void): Promise<AudioNode>;
+}
+
+export { EnergyVADPlugin };