Add app.py and requirements

app.py

@@ -0,0 +1,64 @@
+import gradio as gr
+from TTS.api import TTS
+import os
+
+# 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)
+
+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 a default voice if specific one missing, or error out
+        # For now, let's list available voices to help debug
+        available = os.listdir("voices") if os.path.exists("voices") else []
+        return None, f"Error: Voice ID '{voice_id}' not found. Available: {available}"
+
+    # Generate
+    tts.tts_to_file(
+        text=text,
+        file_path=output_path,
+        speaker_wav=speaker_wav,
+        language="en"
+    )
+    
+    return output_path, "Success"
+
+# 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()

requirements.txt

@@ -0,0 +1,3 @@
+TTS
+torch
+torchaudio