app.py

import gradio as gr
import edge_tts
import io
import os
import time
import asyncio
from docx import Document

AUDIO_DIR = 'audio_files'  # Thư mục lưu tệp âm thanh
MAX_FILE_AGE = 24 * 60 * 60  # Thời gian tối đa lưu trữ tệp âm thanh (24 giờ)

async def text_to_speech(text, lang):
    tts = edge_tts.Communicate(text, voice=lang)
    audio_fp = io.BytesIO()  # Tạo một file âm thanh trong bộ nhớ
    metadata_fp = io.BytesIO()  # Tạo một file metadata trong bộ nhớ

    # Lưu âm thanh vào bộ nhớ
    await tts.save(audio_fp, 'audio/mp3')  # Định dạng mp3
    audio_fp.seek(0)

    os.makedirs(AUDIO_DIR, exist_ok=True)  # Đảm bảo thư mục tồn tại
    audio_file_name = f"{time.time()}.mp3"
    audio_file_path = os.path.join(AUDIO_DIR, audio_file_name)

    # Lưu tệp âm thanh vào thư mục AUDIO_DIR
    with open(audio_file_path, 'wb') as f:
        f.write(audio_fp.read())

    delete_old_audio_files()
    return audio_file_path, audio_file_path  # Trả về đường dẫn tệp âm thanh

def delete_old_audio_files():
    now = time.time()
    for file_name in os.listdir(AUDIO_DIR):
        file_path = os.path.join(AUDIO_DIR, file_name)
        if now - os.path.getmtime(file_path) > MAX_FILE_AGE:
            os.remove(file_path)

async def txt_to_speech(file, lang):
    with open(file.name, 'r') as f:
        text = f.read()
    return await text_to_speech(text, lang)

async def docx_to_speech(file, lang):
    doc = Document(file.name)
    text = "\n".join([para.text for para in doc.paragraphs])  # Lấy tất cả văn bản từ các đoạn
    return await text_to_speech(text, lang)

# Tạo giao diện Gradio
with gr.Blocks() as iface:
    with gr.Tab("Text to Speech"):
        gr.Markdown("### Convert text to speech")
        text_input = gr.Textbox(lines=10, label="Enter your text here:")
        lang_input = gr.Dropdown(choices=["vi-VN-NamMinhNeural", "en-US-JennyNeural"], label="Select language:")
        
        audio_output, file_output = gr.Audio(label="Audio"), gr.File(label="Audio File")
        gr.Button("Convert").click(fn=lambda text, lang: asyncio.run(text_to_speech(text, lang)),
                                    inputs=[text_input, lang_input], 
                                    outputs=[audio_output, file_output])

    with gr.Tab("TXT to Speech"):
        gr.Markdown("### Convert .txt file to speech")
        file_input = gr.File(label="Upload your .txt file")
        lang_input_file = gr.Dropdown(choices=["vi-VN-NamMinhNeural", "en-US-JennyNeural"], label="Select language:")
        
        audio_output_file, file_output_file = gr.Audio(label="Audio"), gr.File(label="Audio File")
        gr.Button("Convert").click(fn=lambda file, lang: asyncio.run(txt_to_speech(file, lang)),
                                    inputs=[file_input, lang_input_file], 
                                    outputs=[audio_output_file, file_output_file])

    with gr.Tab("DOCX to Speech"):
        gr.Markdown("### Convert .docx file to speech")
        docx_file_input = gr.File(label="Upload your .docx file")
        lang_input_docx = gr.Dropdown(choices=["vi-VN-NamMinhNeural", "en-US-JennyNeural"], label="Select language:")
        
        audio_output_docx, file_output_docx = gr.Audio(label="Audio"), gr.File(label="Audio File")
        gr.Button("Convert").click(fn=lambda file, lang: asyncio.run(docx_to_speech(file, lang)),
                                    inputs=[docx_file_input, lang_input_docx], 
                                    outputs=[audio_output_docx, file_output_docx])

iface.launch(enable_queue=True)