app.py

import gradio as gr
import torch
from TTS.api import TTS
import os
import spaces
import tempfile

os.environ["COQUI_TOS_AGREED"] = "1"

device = "cuda" if torch.cuda.is_available() else "cpu"

# Initialize TTS model
def load_tts_model():
    return TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)

tts = load_tts_model()

# Celebrity voices (example list, you may want to expand or modify this)
celebrity_voices = {
    "Morgan Freeman": "./voices/Morgan Freeman.mp3",
    "Scarlett Johansson": "./voices/Scarlett Johansson.mp3",
    "David Attenborough": "./voices/David Attenborough.mp3",
    "Tom Hanks": "./voices/Tom Hanks.mp3",
    "Emma Watson": "./voices/Emma Watson.mp3",
    "Batman": "./voices/Batman.mp3",
    "Spongebob": "./voices/Spongebob.mp3",
    "Darth Vader": "./voices/Darth Vader.mp3",
    "Homer Simpson": "./voices/Homer Simpson.mp3",
    "Mario": "./voices/Mario.mp3",
    "PewDiePie": "./voices/PewDiePie.mp3",
    "Pokimane": "./voices/Pokimane.mp3",
    "Ninja": "./voices/Ninja.mp3",
    "Shroud": "./voices/Shroud.mp3",
    "Tfue": "./voices/Tfue.mp3",
    "Barack Obama": "./voices/Barack Obama.mp3",
    "Donald Trump": "./voices/Donald Trump.mp3",
    "Angela Merkel": "./voices/Angela Merkel.mp3",
    "Justin Trudeau": "./voices/Justin Trudeau.mp3",
    "Emmanuel Macron": "./voices/Emmanuel Macron.mp3",
    "Serena Williams": "./voices/Serena Williams.mp3",
    "Michael Jordan": "./voices/Michael Jordan.mp3",
    "Lionel Messi": "./voices/Lionel Messi.mp3",
    "LeBron James": "./voices/LeBron James.mp3",
    "Usain Bolt": "./voices/Usain.mp3"  # Note: The file is named "Usain.mp3"
}

def check_voice_files():
    """
    Checks if all voice files exist in the celebrity_voices dictionary.
    Returns a message listing missing files or confirming all files are present.
    """
    missing = []
    for voice, path in celebrity_voices.items():
        if not os.path.exists(path):
            missing.append(f"{voice}: {path}")
    if missing:
        return "**Missing Voice Files:**\n" + "\n".join(missing)
    else:
        return "**All voice files are present.** 🎉"

@spaces.GPU(duration=120)
def tts_generate(text, voice, language):
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
        temp_audio_path = temp_audio.name
    
    tts.tts_to_file(
        text=text,
        speaker_wav=celebrity_voices[voice],
        language=language,
        file_path=temp_audio_path
    )
    
    return temp_audio_path
    
@spaces.GPU(enable_queue=True)
def clone_voice(text, audio_file, language):
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
        temp_audio_path = temp_audio.name
    
    tts.tts_to_file(
        text=text, 
        speaker_wav=audio_file,
        language=language,
        file_path=temp_audio_path
    )
    
    return temp_audio_path

# Placeholder function for Talking Image tab
def talking_image_placeholder():
    return "Talking Image functionality not implemented yet."

# Define Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Advanced Voice Synthesis")
    
    # Display voice files status
    voice_status = check_voice_files()
    gr.Markdown(voice_status)
    
    with gr.Tabs():
        with gr.TabItem("TTS"):
            with gr.Row():
                tts_text = gr.Textbox(label="Text to speak")
                tts_voice = gr.Dropdown(choices=list(celebrity_voices.keys()), label="Celebrity Voice")
                tts_language = gr.Dropdown(["en", "es", "fr", "de", "it","ar"], label="Language", value="en")
            tts_generate_btn = gr.Button("Generate")
            tts_output = gr.Audio(label="Generated Audio")
            
            tts_generate_btn.click(
                tts_generate,
                inputs=[tts_text, tts_voice, tts_language],
                outputs=tts_output
            )
        
        with gr.TabItem("Talking Image"):
            gr.Markdown("Talking Image functionality coming soon!")
        
        with gr.TabItem("Clone Voice"):
            with gr.Row():
                clone_text = gr.Textbox(label="Text to speak")
                clone_audio = gr.Audio(label="Voice reference audio file", type="filepath")
                clone_language = gr.Dropdown(["en", "es", "fr", "de", "it","ar"], label="Language", value="en")
            clone_generate_btn = gr.Button("Generate")
            clone_output = gr.Audio(label="Generated Audio")
            
            clone_generate_btn.click(
                clone_voice,
                inputs=[clone_text, clone_audio, clone_language],
                outputs=clone_output
            )

js_func = """
function refresh() {
    const url = new URL(window.location);

    if (url.searchParams.get('__theme') !== 'dark') {
        url.searchParams.set('__theme', 'dark');
        window.location.href = url.href;
    }
}
"""

# Launch the interface
demo.launch()

# Clean up temporary files (this will run after the Gradio server is closed)
for file in os.listdir():
    if file.endswith('.wav') and file.startswith('tmp'):
        os.remove(file)