Update app.py

app.py

@@ -18,10 +18,9 @@ def predict_audio(audio_path):
     global model
     
     if audio_path is None:
-        return {"error": "Please upload an audio file first."}
+        return "Error: Upload file audio dulu.", "-", "-", "-", "-"
     
     try:
-        # Pindah semua import library berat ke dalem fungsi
         import librosa
         import cv2
         import tensorflow as tf
@@ -29,8 +28,14 @@ def predict_audio(audio_path):
         if model is None:
             model = tf.keras.models.load_model(model_path)
             
-        y, sr = librosa.load(audio_path, sr=22050, duration=3.0)
+        # Hitung durasi dan ambil tengahnya
+        total_duration = librosa.get_duration(path=audio_path)
+        offset_time = (total_duration / 2.0) - 1.5 if total_duration > 6.0 else 0.0
+        
+        # Load 3 detik dari tengah
+        y, sr = librosa.load(audio_path, sr=22050, offset=offset_time, duration=3.0)
         y = librosa.util.fix_length(y, size=3*22050)
+        
         spec = librosa.feature.melspectrogram(y=y, sr=sr, n_mels=128)
         spec_db = librosa.power_to_db(spec, ref=np.max)
         img = (spec_db - spec_db.min()) / (spec_db.max() - spec_db.min()) * 255.0
@@ -40,20 +45,29 @@ def predict_audio(audio_path):
 
         preds = model.predict(input_tensor, verbose=0)
         
-        return {
-            "mbti": str(labels['mbti'][np.argmax(preds[0])]),
-            "emotion": str(labels['emotion'][np.argmax(preds[1])]),
-            "vibe": str(labels['vibe'][np.argmax(preds[2])]),
-            "intensity": str(labels['intensity'][np.argmax(preds[3])]),
-            "tempo": str(labels['tempo'][np.argmax(preds[4])])
-        }
+        # Ekstrak hasil prediksi jadi variabel terpisah
+        mbti = str(labels['mbti'][np.argmax(preds[0])])
+        emotion = str(labels['emotion'][np.argmax(preds[1])])
+        vibe = str(labels['vibe'][np.argmax(preds[2])])
+        intensity = str(labels['intensity'][np.argmax(preds[3])])
+        tempo = str(labels['tempo'][np.argmax(preds[4])])
+        
+        return mbti, emotion, vibe, intensity, tempo
+
     except Exception as e:
-        return {"error": str(e)}
+        return f"Error: {str(e)}", "-", "-", "-", "-"
 
+# UI Gradio yang lebih rapi pake Textbox terpisah
 demo = gr.Interface(
     fn=predict_audio,
     inputs=gr.Audio(type="filepath", label="Upload Math Rock / Midwest Emo Audio"),
-    outputs=gr.JSON(label="Classification Results"),
+    outputs=[
+        gr.Textbox(label="MBTI Personality"),
+        gr.Textbox(label="Emotion"),
+        gr.Textbox(label="Audio Vibe"),
+        gr.Textbox(label="Intensity"),
+        gr.Textbox(label="Tempo")
+    ],
     title="Neural Mathrock Classifier"
 )