@vkontakte/calls-audio-effects 1.2.10-dev.12fb226c.0 → 1.2.10-dev.652c81ba.0

package/controllers/VadController.d.ts

@@ -1,7 +1,8 @@
-export interface VadControllerOptions {
-    /** Порог вероятности речи для включения трека. Default: 0.66 */
+import { type EffectVadAiOptions } from '../effects/EffectVadAi';
+export interface VadControllerOptions extends EffectVadAiOptions {
+    /** Speech probability threshold for enabling track. Default: 0.66 */
     threshold?: number;
-    /** Сколько ms после последнего "речь" держим enabled=true. Default: 300 */
+    /** How many ms after last "speech" to keep enabled=true. Default: 300 */
     hangoverMs?: number;
 }
 export interface VadLib {
@@ -13,6 +14,7 @@ export declare class VadController {
     private _vadInputTrack;
     private _threshold;
     private _hangoverMs;
+    private _effectOptions;
     private _isSpeaking;
     private _lastSpeechAt;
     private _hangoverTimer;
@@ -22,6 +24,7 @@ export declare class VadController {
     private _logger;
     constructor(sendTrack?: MediaStreamTrack | null, options?: VadControllerOptions);
     setOptions(options: VadControllerOptions): void;
+    private _applyEffectOptions;
     setLogger(loggerFunction: (type: string, ...args: unknown[]) => void, loggerName?: string): void;
     init(): Promise<void>;
     setSendTrack(track: MediaStreamTrack | null): void;

package/effects/EffectVadAi.d.ts

@@ -1,11 +1,31 @@
 import { DataProcessing } from '../processor/AudioProcessor';
 import { WorkerAudioProcessor } from '../processor/WorkerAudioProcessor';
+export type VadModelType = 'onnx' | 'tflite';
+export declare const DEFAULT_VAD_AI_URLS: {
+    readonly modelUrl: "https://calls-sdk.cdn-vk.ru/models/bcresnet_no_center_mult2_20ms.onnx";
+    readonly modelDataUrl: "https://calls-sdk.cdn-vk.ru/models/bcresnet_no_center_mult2_20ms.onnx.data";
+    readonly ortWasmUrl: "https://calls-sdk.cdn-vk.ru/vendor/onnxruntime-web/ort-wasm-simd-threaded.wasm";
+    readonly ortMjsUrl: "https://calls-sdk.cdn-vk.ru/vendor/onnxruntime-web/ort-wasm-simd-threaded.mjs";
+};
+export interface EffectVadAiOptions {
+    modelType?: VadModelType;
+    modelUrl?: string;
+    modelDataUrl?: string;
+    ortWasmUrl?: string;
+    ortMjsUrl?: string;
+}
 export interface VadProbs {
     speechProb: number;
     silenceProb: number;
 }
 export declare class EffectVadAi extends WorkerAudioProcessor implements DataProcessing {
     onData: (probs: VadProbs) => void;
+    private readonly _modelType;
+    private readonly _modelUrl;
+    private readonly _modelDataUrl;
+    private readonly _ortWasmUrl;
+    private readonly _ortMjsUrl;
+    constructor(options?: EffectVadAiOptions);
     init(): Promise<void>;
     protected onWorkletData(data: ArrayBuffer): void;
     protected onWorkerData(probs: VadProbs): void;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vkontakte/calls-audio-effects",
-  "version": "1.2.10-dev.12fb226c.0",
+  "version": "1.2.10-dev.652c81ba.0",
   "author": "vk.com",
   "description": "Audio effects library for video calls based on the vk.com platform",
   "homepage": "https://vk.com",
@@ -14,7 +14,7 @@
     "**/*.d.ts"
   ],
   "dependencies": {
-    "@vkontakte/audio-processing": "0.0.8",
+    "@vkontakte/audio-processing": "0.0.8-dev.0848d3fd",
     "onnxruntime-web": "^1.26.0"
   }
 }

package/workers/VadTfliteWorker.d.ts

@@ -0,0 +1,6 @@
+import wav2lip from '@vkontakte/audio-processing';
+export interface VadWorkerRequest {
+    data: ArrayBuffer;
+}
+declare const _default: (createWav2lipModule: typeof wav2lip, urlResolver: (url: string) => string) => void;
+export default _default;