@valoric/whisper-node/dist/tsToArray.js

Local audio transcription on CPU. Node.js bindings for OpenAI's Whisper.

"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
function parseTranscript(vtt) {
    // 1. separate lines by matching the format like "[00:03:04.000 --> 00:03:13.000]   XXXXXX"
    const lines = vtt.match(/\[[0-9:.]+\s-->\s[0-9:.]+\].*/g);
    // 2. convert each line into an object
    return lines.map(line => {
        // 2a. split ts from speech
        let [timestamp, speech] = line.split(']  '); // two spaces (3 spaces doesn't work with punctuation like period . )
        // 2b. remove the open bracket of timestamp
        timestamp = timestamp.substring(1);
        // 2c. split timestamp into begin and end
        const [start, end] = timestamp.split(' --> ');
        // 2d. remove \n from speech with regex
        speech = speech.replace(/\n/g, '');
        // 2e. remove beginning space
        speech = speech.replace(' ', '');
        return { start, end, speech };
    })
        // 3. filter out empty transcriptions
        .filter(({ speech }) => speech.length);
}
exports.default = parseTranscript;
//# sourceMappingURL=tsToArray.js.map