generate_speech_samples.py

import os
import requests
from tqdm import tqdm

MALE_VOICE = "Andrew"
FEMALE_VOICE = "Ava"

TEXT_TO_SPEECH_URL = "http://localhost:5379"
TEXT_TO_SPEECH_ENDPOINT = f"{TEXT_TO_SPEECH_URL}/text_to_speech"

SPEECH_SAMPLE_DIR = "shared/input"


def generate_speech_samples():
    speeches = [
        "The quick brown fox jumps over the lazy dog.",
        "Where do bright violets bloom in relation to ancient oaks?",
        # "Complex systems evolve from simple rules.",
        # "Thunderstorms could delay the early morning flight.",
        "A murmuration of starlings filled the dusky sky.",
        # "She sells seashells by the seashore.",
        # "The mysterious manuscript contained unfamiliar symbols.",
        # "Artificial intelligence transforms industries globally.",
        "What type of melody did the jazz saxophonist play?",
        # "Climate change affects global weather patterns.",
        "Quantum computing revolutionizes data processing, offering solutions to previously unsolvable problems.",
        # "Digital encryption secures online transactions, right?",
        # "Ancient civilizations crafted remarkable artifacts.",
        # "The culinary artist designed a five-course meal.",
        # "How do renewable energy sources impact carbon emissions?",
        # "Theoretical physics explores multidimensional spaces.",
        # "Quantum computing revolutionizes data processing.",
        "The ancient library contained manuscripts that chronicled the lost civilization's history and knowledge.",
        # "Bioluminescent creatures light up the ocean depths.",
        # "The geneticist decoded the DNA sequence efficiently.",
    ]

    male_voice_lowercase = MALE_VOICE.lower()
    male_output_dir = os.path.join(SPEECH_SAMPLE_DIR, male_voice_lowercase)

    female_voice_lowercase = FEMALE_VOICE.lower()
    female_output_dir = os.path.join(SPEECH_SAMPLE_DIR, female_voice_lowercase)
    os.makedirs(male_output_dir, exist_ok=True)
    os.makedirs(female_output_dir, exist_ok=True)

    for index, speech in tqdm(enumerate(speeches)):
        body = {"voice": MALE_VOICE, "inputText": speech}

        response = requests.post(TEXT_TO_SPEECH_ENDPOINT, json=body)
        file_path = os.path.join(male_output_dir, f"{index}.wav")
        with open(file_path, "wb") as f:
            f.write(response.content)

    for index, speech in tqdm(enumerate(speeches)):
        body = {"voice": FEMALE_VOICE, "inputText": speech}

        response = requests.post(TEXT_TO_SPEECH_ENDPOINT, json=body)
        file_path = os.path.join(female_output_dir, f"{index}.wav")
        with open(file_path, "wb") as f:
            f.write(response.content)


if __name__ == "__main__":
    generate_speech_samples()