app.py

import logging
from pdfminer.high_level import extract_text
from huggingface_hub import hf_hub_download
from balacoon_tts import TTS
import gradio as gr

# Global TTS module, initialized from a selected model
tts = None

def read_pdf(file):
    text = extract_text(file.name)
    return text

def synthesize_audio(file):
    text_str = read_pdf(file)
    if len(text_str) > 1024:
        text_str = text_str[:1024]
    samples = tts.synthesize(text_str)
    return (samples, tts.get_sampling_rate())

def main():
    logging.basicConfig(level=logging.INFO)
    file_input = gr.inputs.File(label="Upload PDF")

    model_name = "balacoon/tts"  # Set the desired model name here
    model_path = hf_hub_download(repo_id=model_name)
    global tts
    tts = TTS(model_path)

    audio = gr.outputs.Audio(label="Generated Audio", type="numpy")

    def generate_audio(file):
        return synthesize_audio(file)

    iface = gr.Interface(
        fn=generate_audio,
        inputs=file_input,
        outputs=audio,
        title="PDF TO SPEECH CONVERTER",
        layout="vertical",
        debug=True
    )

    iface.launch()

if __name__ == "__main__":
    main()