@picovoice/cheetah-react-native/lib/typescript/cheetah_types.d.ts

Picovoice Cheetah React Native binding

export type CheetahTranscript = {
    /** Any newly-transcribed speech. If none is available then an empty string is returned. */
    transcript: string;
    /** Flag indicating if an endpoint has been detected. */
    isEndpoint?: boolean;
};
export type CheetahOptions = {
    /** String representation of the device (e.g., CPU or GPU) to use for inference.
     * If set to `best`, the most suitable device is selected automatically. If set to `gpu`, the engine uses the
     * first available GPU device. To select a specific GPU device, set this argument to `gpu:${GPU_INDEX}`, where
     * `${GPU_INDEX}` is the index of the target GPU. If set to `cpu`, the engine will run on the CPU with the
     * default number of threads. To specify the number of threads, set this argument to `cpu:${NUM_THREADS}`,
     * where `${NUM_THREADS}` is the desired number of threads.
     */
    device?: string;
    /** Duration of endpoint in seconds. A speech endpoint is detected when there is a segment
     * of audio (with a duration specified herein) after an utterance without any speech in it. Set to `0` to disable
     * endpoint detection.*/
    endpointDuration?: number;
    /** Set to `true` to enable automatic punctuation insertion. */
    enableAutomaticPunctuation?: boolean;
    /** Set to `true` to enable text normalization. Enabling this feature
     * improves the readability and formatting of Cheetah's transcriptions (e.g. converts number words to digits)
     * at the cost of some additional latency. */
    enableTextNormalization?: boolean;
};