Fix UI clear behavior and improve audio UX

app.py

@@ -12,6 +12,9 @@ from audio_backend import predict_audio
 def analyze_image(image):
     label, confidence, heatmap = predict_image_pil(image)
 
+    if image is None:
+        return "", "", "", None
+
     if label == "Fake":
         if confidence >= 90:
             risk = '<span class="material-icons">error</span> High likelihood of deepfake'
@@ -37,9 +40,17 @@ def analyze_audio(audio_path):
     label, confidence, spec_img, error = predict_audio(audio_path)
 
     # ---------- Error handling ----------
+    if audio_path is None:
+        return (
+            "No Input",
+            "-",
+            '<span class="material-icons">warning</span> Please upload an audio file.',
+            None
+        )
+
     if error is not None:
         return (
-            "Error",
+            "Invalid Input",
             "-",
             f'<span class="material-icons">error</span> {error}',
             None
@@ -145,7 +156,7 @@ with gr.Blocks(css="style.css") as demo:
                 with gr.Column(scale=2):
                     img_pred = gr.Text(label="Prediction")
                     img_conf = gr.Text(label="Confidence")
-                    img_risk = gr.HTML(label="Risk Assessment")
+                    img_risk = gr.HTML(label="Risk Assessment", value="")
                     img_heatmap = gr.Image(
                         label="Explainability Heatmap",
                         height=280
@@ -158,9 +169,9 @@ with gr.Blocks(css="style.css") as demo:
             )
 
             img_clear.click(
-                fn=lambda: (None, "", "", None),
+                fn=lambda: (None, "", "", "", None),
                 inputs=None,
-                outputs=[image_input, img_pred, img_conf, img_risk]
+                outputs=[image_input, img_pred, img_conf, img_risk, img_heatmap]
             )
 
         # =========================
@@ -172,17 +183,18 @@ with gr.Blocks(css="style.css") as demo:
             with gr.Row():
                 with gr.Column(scale=1):
                     audio_input = gr.Audio(
-                        label="Upload Audio (.wav)",
+                        label="Upload Audio (WAV, MP3, FLAC, M4A, OGG)",
                         type="filepath"
                     )
+
                     aud_submit = gr.Button("Submit")
                     aud_clear = gr.Button("Clear")
 
                 with gr.Column(scale=2):
                     aud_pred = gr.Text(label="Prediction")
                     aud_conf = gr.Text(label="Confidence")
-                    aud_risk = gr.HTML(label="Risk Assessment")
-                    aud_spec = gr.Image(label="Audio Spectrogram (Model Input)",height=280)
+                    aud_risk = gr.HTML(label="Risk Assessment", value="")
+                    aud_spec = gr.Image(label="Audio Spectrogram (Model Input)",height=280, value="")
 
             aud_submit.click(
                 fn=analyze_audio,
@@ -191,7 +203,7 @@ with gr.Blocks(css="style.css") as demo:
             )
 
             aud_clear.click(
-                fn=lambda: (None, "", ""),
+                fn=lambda: (None, "", "", "", None),
                 inputs=None,
                 outputs=[audio_input, aud_pred, aud_conf, aud_risk, aud_spec]
             )