@ericedouard/vad-node-realtime/dist/real-time-vad.d.ts

Powerful, user-friendly realtime voice activity detector (VAD) for node

import type { OrtOptions } from "./common";
import { type FrameProcessorOptions } from "./common/frame-processor";
import { type SpeechProbabilities } from "./common/models";
export type ModelVersion = "v5" | "legacy";
export declare const DEFAULT_MODEL: ModelVersion;
/**
 * Callbacks for real-time VAD events
 */
export interface RealTimeVADCallbacks {
    onFrameProcessed: (probabilities: SpeechProbabilities, frame: Float32Array) => void;
    onVADMisfire: () => void;
    onSpeechStart: () => void;
    onSpeechRealStart: () => void;
    onSpeechEnd: (audio: Float32Array) => void;
}
/**
 * Options for RealTimeVAD in Node environment
 */
export interface RealTimeVADOptions extends FrameProcessorOptions, RealTimeVADCallbacks, OrtOptions {
    /** Sample rate of the incoming audio; will be resampled to 16000Hz internally */
    sampleRate: number;
    /** Which Silero model to use: V5 or legacy */
    model?: ModelVersion;
}
/**
 * Build default options based on chosen model
 */
export declare function getDefaultRealTimeVADOptions(model?: ModelVersion): RealTimeVADOptions;
/**
 * RealTimeVAD processes raw audio buffers, frames, and emits events
 */
export declare class RealTimeVAD {
    private options;
    private frameProcessor;
    private modelInstance;
    private buffer;
    private frameSize;
    private active;
    private resampler;
    /**
     * Construct a new instance with provided options and loaded model
     */
    protected constructor(options: RealTimeVADOptions, modelInstance: any);
    /**
     * Create and initialize a RealTimeVAD instance
     */
    static new(ort: any, modelFetcher: () => Promise<ArrayBuffer>, opts?: Partial<RealTimeVADOptions>): Promise<RealTimeVAD>;
    /** Start processing incoming frames */
    start(): void;
    /** Pause processing; may emit end-segment on pause */
    pause(): void;
    /** Feed raw audio (any sample rate) into the VAD */
    processAudio(audioData: Float32Array): Promise<void>;
    /** Flush any remaining audio and end segment */
    flush(): Promise<void>;
    /** Reset internal state */
    reset(): void;
    /** Handle events emitted by the frame processor */
    private handleEvent;
    /** Clean up resources */
    destroy(): void;
}
//# sourceMappingURL=real-time-vad.d.ts.map