audiobook.py

import os
import nltk
nltk.download('punkt')
import numpy as np

from bark.generation import (
    generate_text_semantic,
    preload_models,
)
from bark.api import semantic_to_waveform
from bark import generate_audio, SAMPLE_RATE
import soundfile as sf
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

# Loads the model, should be run one time
preload_models()

class AudioBook:
    def __init__(self, output_folder="output"):
        self.output_folder = output_folder

        # Create the output folder if it doesn't exist
        if not os.path.exists(output_folder):
            os.makedirs(output_folder)

    def generate_audio_from_text(self, text, speaker="male", filename="output_audio"):
        # Preprocess text
        text = text.replace("\n", " ").strip()
        sentences = nltk.sent_tokenize(text)

        # Choose the speaker based on the input
        if speaker == "male":
            SPEAKER = "v2/en_speaker_6"
        elif speaker == "female":
            SPEAKER = "v2/en_speaker_9"
        else:
            raise ValueError("Invalid speaker selection. Use 'male' or 'female'.")

        silence = np.zeros(int(0.25 * SAMPLE_RATE))  # quarter-second of silence

        pieces = []
        for sentence in sentences:
            audio_array = generate_audio(sentence, history_prompt=SPEAKER, text_temp=0.7, waveform_temp=0.7)
            pieces += [audio_array, silence.copy()]

        audio_data = np.concatenate(pieces)

        # Save the audio to a WAV file in the output folder
        output_path = os.path.join(self.output_folder, f"{filename}.wav")
        sf.write(output_path, audio_data, SAMPLE_RATE)

        return output_path