Upload app.py with huggingface_hub

app.py

@@ -33,6 +33,18 @@ from transformers import (
     MimiModel,
 )
 
+# Import spaces for GPU support
+try:
+    import spaces
+    GPU_AVAILABLE = True
+except ImportError:
+    GPU_AVAILABLE = False
+    # Create dummy decorator if spaces not available
+    class spaces:
+        @staticmethod
+        def GPU(func):
+            return func
+
 
 # Constants
 SAMPLE_RATE = 24000
@@ -241,6 +253,7 @@ def generate_tts_audio(text: str, voice: str, instructions: str = None) -> str:
         return temp_file.name
 
 
+@spaces.GPU
 def process_inputs(transcript_text: str, voice1: str, voice2: str, instructions1: str = "", instructions2: str = ""):
     """Process inputs and generate audio."""
     if not all([model, mimi_model, tokenizer, feature_extractor, openai_client]):
@@ -318,6 +331,7 @@ def process_inputs(transcript_text: str, voice1: str, voice2: str, instructions1
     )
 
 
+@spaces.GPU
 def swap_embeddings(embeddings_json: str, swap_indices: str):
     """Perform embedding swap at specified token indices."""
     if not embeddings_json: