tts-narrator/dist/tts-service.js

Generate narration with Text-To-Speech technology

"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
exports.BaseTtsService = exports.TtsServiceType = void 0;
const fast_xml_parser_1 = require("fast-xml-parser");
var TtsServiceType;
(function (TtsServiceType) {
    TtsServiceType["Azure"] = "azure";
})(TtsServiceType || (exports.TtsServiceType = TtsServiceType = {}));
function escapeXml(text) {
    return text.replaceAll(/["&'<>]/g, (c) => {
        switch (c) {
            case '<': {
                return '&lt;';
            }
            case '>': {
                return '&gt;';
            }
            case '&': {
                return '&amp;';
            }
            case '\'': {
                return '&apos;';
            }
            case '"': {
                return '&quot;';
            }
            default: {
                return c;
            }
        }
    });
}
class BaseTtsService {
    async generateSSML(paragraph) {
        const generated = this.generateSsmlWithoutValidation(paragraph);
        this.validateXML(generated.ssml, generated.lineOffset);
        return generated.ssml;
    }
    validateXML(xml, lineOffset) {
        const validationResult = fast_xml_parser_1.XMLValidator.validate(xml);
        if (validationResult !== true) {
            const err = validationResult.err;
            throw new Error(`Invalid markup at line ${err.line - lineOffset}. ${err.code}: ${err.msg}`);
        }
    }
    buildSpeakStartTag(voiceSettings) {
        var _a;
        return `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="${(_a = voiceSettings.language) !== null && _a !== void 0 ? _a : 'en-US'}">`;
    }
    buildVoiceStartTag(voiceSettings) {
        return `<voice name="${voiceSettings.name}">`;
    }
    buildProsodyStartTag(prosodySettings) {
        return `<prosody ${prosodySettings.pitch ? `pitch="${prosodySettings.pitch}"` : ''} ${prosodySettings.rate ? `rate="${prosodySettings.rate}"` : ''} ${prosodySettings.volume ? `volume="${prosodySettings.volume}"` : ''}>`;
    }
    buildMsttsExpressAsStartTag(msttsExpressAsSettings) {
        return `<mstts:express-as ${msttsExpressAsSettings.style ? `style="${msttsExpressAsSettings.style}"` : ''} ${msttsExpressAsSettings.role ? `role="${msttsExpressAsSettings.role}"` : ''} ${msttsExpressAsSettings.styleDegree ? `styledegree="${msttsExpressAsSettings.styleDegree}"` : ''}>`;
    }
    generateSsmlWithoutValidation(paragraph) {
        var _a, _b;
        const text = (_a = paragraph === null || paragraph === void 0 ? void 0 : paragraph.text) === null || _a === void 0 ? void 0 : _a.trim();
        const ssml = (_b = paragraph === null || paragraph === void 0 ? void 0 : paragraph.ssml) === null || _b === void 0 ? void 0 : _b.trim();
        if (!text && !ssml) {
            throw new Error('Empty content?');
        }
        if (ssml && ssml.startsWith('<speak')) { // it is already full SSML
            if (!ssml.endsWith('</speak>')) {
                throw new Error('Forgot to end the text with "</speak>"?');
            }
            return { lineOffset: 0, ssml };
        }
        const voiceSettings = paragraph.settings;
        const speakStartTag = this.buildSpeakStartTag(voiceSettings);
        const speakEndTag = '</speak>';
        // <voice></voice> fragment
        if (ssml && ssml.startsWith('<voice')) {
            if (!ssml.endsWith('</voice>')) {
                throw new Error('Forgot to end the text with "</voice>"?');
            }
            return { lineOffset: 1, ssml: `${speakStartTag}\n${ssml}\n${speakEndTag}` };
        }
        const voiceStartTag = this.buildVoiceStartTag(voiceSettings);
        const voiceEndTag = '</voice>';
        let prosodyStartTagOrEmpty = '';
        let prosodyEndTagOrEmpty = '';
        if (voiceSettings.prosody) {
            prosodyStartTagOrEmpty = this.buildProsodyStartTag(voiceSettings.prosody);
            prosodyEndTagOrEmpty = '</prosody>';
        }
        let msttsExpressAsStartTagOrEmpty = '';
        let msttsExpressAsEndTagOrEmpty = '';
        if (voiceSettings.msttsExpressAs) {
            msttsExpressAsStartTagOrEmpty = this.buildMsttsExpressAsStartTag(voiceSettings.msttsExpressAs);
            msttsExpressAsEndTagOrEmpty = '</mstts:express-as>';
        }
        // plain text or fragments containing other tags
        return { lineOffset: 1, ssml: `${speakStartTag}${voiceStartTag}${prosodyStartTagOrEmpty}${msttsExpressAsStartTagOrEmpty}\n${ssml || escapeXml(text)}\n${msttsExpressAsEndTagOrEmpty}${prosodyEndTagOrEmpty}${voiceEndTag}${speakEndTag}` };
    }
    generateAudio(_ssml, _options) {
        throw new Error('generateAudio(...) is not implemented.');
    }
}
exports.BaseTtsService = BaseTtsService;