Upload app.py with huggingface_hub

app.py

@@ -9,6 +9,8 @@ import os
 import sys
 import pickle
 import tempfile
+import numpy as np
+import scipy.io.wavfile as wavfile
 from huggingface_hub import hf_hub_download
 
 # Add text2midi model to path
@@ -124,52 +126,66 @@ def midi_to_wav(midi_path: str, wav_path: str, sample_rate: int = 44100) -> bool
     return os.path.exists(wav_path)
 
 def generate_music(prompt: str):
-    """Generate music from text prompt. Returns audio file path."""
+    """Generate music from text prompt. Returns audio as (sample_rate, numpy_array)."""
     if not prompt or not prompt.strip():
         return None
 
+    midi_path = None
+    wav_path = None
+
     try:
         # Create temporary files
-        midi_file = tempfile.NamedTemporaryFile(suffix='.mid', delete=False)
-        midi_path = midi_file.name
-        midi_file.close()
-
-        wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=False)
-        wav_path = wav_file.name
-        wav_file.close()
-
-        try:
-            # Generate MIDI
-            if MODEL_LOADED:
-                generate_midi_with_model(prompt, midi_path, max_len=512, temperature=0.9)
-            else:
-                from midiutil import MIDIFile
-                midi = MIDIFile(1)
-                midi.addTempo(0, 0, 120)
-                notes = [60, 62, 64, 65, 67, 69, 71, 72]
-                for i, note in enumerate(notes[:min(len(prompt.split()), 8)]):
-                    midi.addNote(0, 0, note, i, 1, 100)
-                with open(midi_path, "wb") as f:
-                    midi.writeFile(f)
-
-            # Convert to WAV
-            if SOUNDFONT_PATH and midi_to_wav(midi_path, wav_path):
-                return wav_path
-            else:
-                return None
-
-        finally:
-            try:
-                os.unlink(midi_path)
-            except:
-                pass
+        midi_fd, midi_path = tempfile.mkstemp(suffix='.mid')
+        os.close(midi_fd)
+
+        wav_fd, wav_path = tempfile.mkstemp(suffix='.wav')
+        os.close(wav_fd)
+
+        # Generate MIDI
+        if MODEL_LOADED:
+            generate_midi_with_model(prompt, midi_path, max_len=512, temperature=0.9)
+        else:
+            from midiutil import MIDIFile
+            midi = MIDIFile(1)
+            midi.addTempo(0, 0, 120)
+            notes = [60, 62, 64, 65, 67, 69, 71, 72]
+            for i, note in enumerate(notes[:min(len(prompt.split()), 8)]):
+                midi.addNote(0, 0, note, i, 1, 100)
+            with open(midi_path, "wb") as f:
+                midi.writeFile(f)
+
+        # Convert to WAV
+        if SOUNDFONT_PATH and midi_to_wav(midi_path, wav_path):
+            # 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
 
     except Exception as e:
         import traceback
         traceback.print_exc()
         return None
 
-# Create simple Gradio Interface (avoids schema generation bugs in gr.Blocks)
+    finally:
+        # Clean up temp files
+        if midi_path and os.path.exists(midi_path):
+            try:
+                os.unlink(midi_path)
+            except:
+                pass
+        if wav_path and os.path.exists(wav_path):
+            try:
+                os.unlink(wav_path)
+            except:
+                pass
+
+# Create simple Gradio Interface
 demo = gr.Interface(
     fn=generate_music,
     inputs=gr.Textbox(
@@ -177,7 +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=[