75 lines
3.0 KiB
JavaScript
75 lines
3.0 KiB
JavaScript
|
"use strict";
|
||
|
|
var __importDefault = (this && this.__importDefault) || function (mod) {
|
||
|
|
return (mod && mod.__esModule) ? mod : { "default": mod };
|
||
|
|
};
|
||
|
|
Object.defineProperty(exports, "__esModule", { value: true });
|
||
|
|
exports.OpenAITTSProvider = void 0;
|
||
|
|
const fs_1 = __importDefault(require("fs"));
|
||
|
|
const child_process_1 = require("child_process");
|
||
|
|
const openai_1 = require("openai");
|
||
|
|
const mediaUtils_1 = require("../../utils/mediaUtils");
|
||
|
|
/**
|
||
|
|
* OpenAI TTS Provider Implementation
|
||
|
|
*/
|
||
|
|
class OpenAITTSProvider {
|
||
|
|
constructor(config) {
|
||
|
|
this.config = config;
|
||
|
|
this.openai = new openai_1.OpenAI({
|
||
|
|
apiKey: config.apiKey,
|
||
|
|
});
|
||
|
|
}
|
||
|
|
/**
|
||
|
|
* Convert text to speech
|
||
|
|
* @param text - Text to convert to speech
|
||
|
|
* @param outputPath - Output path for the audio file
|
||
|
|
* @param options - Additional options
|
||
|
|
* @returns Duration of the generated audio in seconds and cost
|
||
|
|
*/
|
||
|
|
async textToSpeech(text, outputPath, options = {}) {
|
||
|
|
try {
|
||
|
|
// Get the options, with defaults from config
|
||
|
|
const voice = options.voice || this.config.voice;
|
||
|
|
const model = options.model || this.config.model;
|
||
|
|
const speedFactor = options.speedFactor || 1.0;
|
||
|
|
// Generate the initial TTS output
|
||
|
|
const tempOutputPath = outputPath.replace(/\.\w+$/, '_temp$&');
|
||
|
|
const mp3 = await this.openai.audio.speech.create({
|
||
|
|
model: model,
|
||
|
|
voice: voice,
|
||
|
|
input: text,
|
||
|
|
...(options.instructions ? { instructions: options.instructions } : {})
|
||
|
|
});
|
||
|
|
// Cost calculation is based on character count
|
||
|
|
const cost = text.length;
|
||
|
|
const buffer = Buffer.from(await mp3.arrayBuffer());
|
||
|
|
fs_1.default.writeFileSync(tempOutputPath, buffer);
|
||
|
|
// Speed up the audio using FFmpeg if needed
|
||
|
|
if (speedFactor !== 1.0) {
|
||
|
|
(0, child_process_1.execSync)(`ffmpeg -v error -i "${tempOutputPath}" -filter:a "atempo=${speedFactor}" -c:a libmp3lame -q:a 2 "${outputPath}" -y`);
|
||
|
|
// Clean up temporary file
|
||
|
|
fs_1.default.unlinkSync(tempOutputPath);
|
||
|
|
}
|
||
|
|
else {
|
||
|
|
// Just use the file as is
|
||
|
|
fs_1.default.renameSync(tempOutputPath, outputPath);
|
||
|
|
}
|
||
|
|
// Get actual audio duration for accurate timing
|
||
|
|
const audioDuration = (0, mediaUtils_1.getAudioDuration)(outputPath);
|
||
|
|
return {
|
||
|
|
duration: audioDuration,
|
||
|
|
cost: cost
|
||
|
|
};
|
||
|
|
}
|
||
|
|
catch (error) {
|
||
|
|
console.error("Error generating speech:", error);
|
||
|
|
// Create a silent audio file if TTS fails
|
||
|
|
(0, child_process_1.execSync)(`ffmpeg -v error -f lavfi -i anullsrc=r=24000:cl=mono -t 1 -q:a 9 -acodec libmp3lame "${outputPath}" -y`);
|
||
|
|
return {
|
||
|
|
duration: 1,
|
||
|
|
cost: 0
|
||
|
|
};
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
exports.OpenAITTSProvider = OpenAITTSProvider;
|
||
|
|
//# sourceMappingURL=openAITTSProvider.js.map
|