Upload app.py with huggingface_hub

app.py

@@ -9,7 +9,8 @@ import os
 import sys
 import pickle
 import tempfile
-import base64
+import numpy as np
+import scipy.io.wavfile as wavfile
 from huggingface_hub import hf_hub_download
 
 # Add text2midi model to path
@@ -155,19 +156,14 @@ def generate_music(prompt: str):
 
         # Convert to WAV
         if SOUNDFONT_PATH and midi_to_wav(midi_path, wav_path):
-            # Read WAV file as bytes and encode as base64
-            with open(wav_path, 'rb') as f:
-                audio_bytes = f.read()
-
-            audio_base64 = base64.b64encode(audio_bytes).decode('utf-8')
-
-            # Return as dict with base64 data - this format works with Gradio's Audio component
-            # and doesn't require file download from the client
-            return {
-                "name": "generated_audio.wav",
-                "data": f"data:audio/wav;base64,{audio_base64}",
-                "is_file": False
-            }
+            # Read WAV file and return as numpy array
+            sample_rate, audio_data = wavfile.read(wav_path)
+            # Convert to float32 for Gradio compatibility
+            if audio_data.dtype == np.int16:
+                audio_data = audio_data.astype(np.float32) / 32768.0
+            elif audio_data.dtype == np.int32:
+                audio_data = audio_data.astype(np.float32) / 2147483648.0
+            return (sample_rate, audio_data)
         else:
             return None
 
@@ -197,7 +193,7 @@ demo = gr.Interface(
         placeholder="A cheerful pop song with piano and drums in C major",
         lines=2
     ),
-    outputs=gr.Audio(label="Generated Music", type="filepath"),
+    outputs=gr.Audio(label="Generated Music", type="numpy"),
     title="VR Game Music Generator",
     description="Generate music from text descriptions using AI. Enter a prompt describing the music you want.",
     examples=[