app.py

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

# Agree to Coqui TOS
os.environ["COQUI_TOS_AGREED"] = "1"

# Initialize TTS (XTTS-v2)
# This will download the model on the first run
# We use gpu=True if available, otherwise False
use_gpu = True
try:
    import torch
    if not torch.cuda.is_available():
        use_gpu = False
except:
    use_gpu = False

print(f"Initializing TTS with GPU={use_gpu}...")
tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=use_gpu)

# Output folder
os.makedirs("output", exist_ok=True)
os.makedirs("voices", exist_ok=True)

# Ensure at least one voice exists
if not os.listdir("voices"):
    print("No voices found. Downloading sample...")
    try:
        # Download a sample female voice
        urllib.request.urlretrieve("https://huggingface.co/spaces/coqui/xtts/resolve/main/examples/female.wav", "voices/female_calm.wav")
        print("Downloaded female_calm.wav")
    except Exception as e:
        print(f"Failed to download sample voice: {e}")

def generate_speech(text, voice_id):
    """
    Generates speech from text using a specific voice clone.
    voice_id: The name of the wav file in 'voices/' folder to clone.
    """
    output_path = "output/output.wav"
    
    # Map voice_id to a sample file
    # You must upload these files to your Space's 'voices/' folder
    speaker_wav = f"voices/{voice_id}.wav"
    
    if not os.path.exists(speaker_wav):
        # Fallback to the first available voice if specific one missing
        available = os.listdir("voices")
        if available:
            print(f"Voice '{voice_id}' not found. Falling back to '{available[0]}'")
            speaker_wav = f"voices/{available[0]}"
        else:
            return None, f"Error: Voice ID '{voice_id}' not found and no voices available."

    # Generate
    try:
        tts.tts_to_file(
            text=text,
            file_path=output_path,
            speaker_wav=speaker_wav,
            language="en"
        )
        return output_path, "Success"
    except Exception as e:
        return None, f"Error generating speech: {str(e)}"

# Define Gradio Interface
iface = gr.Interface(
    fn=generate_speech,
    inputs=[
        gr.Textbox(label="Text"),
        gr.Textbox(label="Voice ID (e.g. 'narrator', 'aribeth')")
    ],
    outputs=[
        gr.Audio(label="Generated Audio", type="filepath"),
        gr.Textbox(label="Status")
    ],
    title="Unreal Engine TTS Backend (XTTS-v2)",
    description="API for DGG_001 Game. Send POST requests to /run/predict"
)

iface.launch()