import spaces
import gradio as gr
import torch
from TTS.api import TTS
import os
os.environ["COQUI_TOS_AGREED"] = "1"
import gradio as gr
import torch
from TTS.api import TTS
import spaces
import os

# 1. Initialize the model on CPU first
# We do this OUTSIDE the function so it doesn't reload every time you click the button
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)

@spaces.GPU(duration=60) # Give it enough time for voice cloning
def clone(text, audio):
    # 2. Check if CUDA is actually available before moving
    if torch.cuda.is_available():
        tts.to("cuda")
    
    # 3. Perform the synthesis
    tts.tts_to_file(
        text=text, 
        speaker_wav=audio, 
        language="en", 
        file_path="./output.wav"
    )
    
    # 4. Move it back to CPU to be a good citizen on ZeroGPU
    if torch.cuda.is_available():
        tts.to("cpu")
        
    return "./output.wav"

# ... your Gradio Blocks/Interface code ...

iface = gr.Interface(fn=clone, 
                     inputs=[gr.Textbox(label='Text'),gr.Audio(type='filepath', label='Voice reference audio file')], 
                     outputs=gr.Audio(type='filepath'),
                     title='Voice Clone',
                     description="""
                     by [Tony Assi](https://www.tonyassi.com/)

                     ---

                     <h3>If you like voice clone then try <a href="https://huggingface.co/spaces/tonyassi/video-face-swap" target="_blank" rel="noopener noreferrer">Video Face Swap</a></h3>

                     ---
                    
                     This space uses xtts_v2 model. Non-commercial use only. [Coqui Public Model License](https://huggingface.co/coqui/XTTS-v2/blob/main/LICENSE.txt)
                     
                     Please ❤️ this Space. [Email me](mailto:tony.assi.media@gmail.com).
                     """,
                     theme = gr.themes.Base(primary_hue="teal",secondary_hue="teal",neutral_hue="slate"),
                     examples=[["Hey! It's me Dorthy, from the Wizard of Oz. Type in whatever you'd like me to say.","./audio/Wizard-of-Oz-Dorthy.wav"],
                               ["It's me Vito Corleone, from the Godfather. Type in whatever you'd like me to say.","./audio/Godfather.wav"],
                               ["Hey, it's me Paris Hilton. Type in whatever you'd like me to say.","./audio/Paris-Hilton.mp3"],
                               ["Hey, it's me Megan Fox from Transformers. Type in whatever you'd like me to say.","./audio/Megan-Fox.mp3"],
                               ["Hey there, it's me Jeff Goldblum. Type in whatever you'd like me to say.","./audio/Jeff-Goldblum.mp3"],
                               ["Hey there, it's me Heath Ledger as the Joker. Type in whatever you'd like me to say.","./audio/Heath-Ledger.mp3"],])
iface.launch()