import type * as ort from "onnxruntime-web";
export type AudioSamples = Float32Array | number[] | readonly number[];
/** Frame-aligned token timing emitted by a decoder. */
export interface TokenFrame {
    startFrame: number;
    endFrame: number;
}
/** Word-level timestamp in seconds. */
export interface WordTimestamp {
    word: string;
    start: number;
    end: number;
}
/** Speech segment timestamp in both samples and seconds. */
export interface SegmentTimestamp {
    start: number;
    end: number;
    startSec: number;
    endSec: number;
}
/** Normalized transcription result returned by ASR models. */
export interface TranscriptResult {
    text: string;
    tokenIds: number[];
    tokenFrames: TokenFrame[];
    words: WordTimestamp[];
    segments?: SegmentTimestamp[];
}
/** Raw VAD probability trace over processed windows. */
export interface VadSpeechProbabilities {
    probs: number[];
    processedSamples: number;
}
/** Runtime tuning knobs for VAD chunking. */
export interface VadRuntimeOptions {
    sampleRate?: number;
    threshold?: number;
    negThreshold?: number;
    minSpeechMs?: number;
    minSilenceMs?: number;
    speechPadMs?: number;
    windowSamples?: number;
}
/** Common ASR transcription interface shared by Node and browser loaders. */
export interface AsrTranscriber {
    sampleRate: number;
    transcribeSamples(samples: AudioSamples, sampleRate?: number, options?: Record<string, unknown>): Promise<TranscriptResult>;
    transcribeWavBuffer(arrayBuffer: ArrayBuffer, options?: Record<string, unknown>): Promise<TranscriptResult>;
}
/** Voice activity detector contract used by `withVadModel()`. */
export interface VadDetector {
    sampleRate: number;
    detectSpeechSegments(samples: AudioSamples, sampleRate?: number, overrides?: VadRuntimeOptions): Promise<SegmentTimestamp[]>;
    speechProbabilities?(samples: AudioSamples, sampleRate?: number): Promise<VadSpeechProbabilities>;
}
/** Global ONNX Runtime Web environment options. */
export interface ConfigureOrtWebOptions {
    numThreads?: number;
    wasmPaths?: string | Record<string, string>;
    simd?: boolean;
    proxy?: boolean;
}
/** Generic decoder tuning supported by transducer loaders. */
export interface DecoderOptions {
    maxSymbols?: number;
}
export type SessionOptions = ort.InferenceSession.SessionOptions;
export type OrtTensor = ort.Tensor;
export type TensorData = ort.Tensor["data"];
export type TensorMap = ort.InferenceSession.OnnxValueMapType;
/** Shared load options accepted by Node and browser model loaders. */
export interface CommonModelLoadOptions {
    quantization?: "int8" | "none" | string;
    sessionOptions?: SessionOptions;
    decoderOptions?: DecoderOptions;
    sampleRate?: number;
    vadModel?: VadDetector | null;
    vad?: VadDetector | null;
    vadOptions?: VadRuntimeOptions;
}
//# sourceMappingURL=types.d.ts.map