app.py

import os
import gradio as gr
from TTS.api import TTS

# License Agreement
os.environ["COQUI_TOS_AGREED"] = "1"

# Model Load (CPU Mode - GPU False)
print("⏳ Model load ho raha hai... Isme 2-3 minute lagenge...")
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)

def predict(text, speaker_audio):
    output_path = "output.wav"
    # Hindi Generation
    tts.tts_to_file(
        text=text,
        speaker_wav=speaker_audio,
        language="hi",
        file_path=output_path
    )
    return output_path

# Simple Interface
demo = gr.Interface(
    fn=predict,
    inputs=[
        gr.Textbox(label="Text"),
        gr.Audio(label="Reference Audio", type="filepath")
    ],
    outputs=gr.Audio(label="Generated Audio"),
)

demo.launch()