improved UI

app.py

@@ -535,6 +535,7 @@ def build_demo():
                         type="filepath",
                         label="Upload Audio",
                     )
+                    gr.Markdown("<div style='display:flex;align-items:center;justify-content:center;height:100%;font-size:1.2em;font-weight:bold;color:#888;'>OR</div>")
                     audio_mic = gr.Audio(
                         sources=["microphone"],
                         type="filepath",
@@ -544,7 +545,7 @@ def build_demo():
 
             with gr.Column():
                 gallery = gr.Gallery(label="Generated Images")
-                status = gr.Markdown(visible=False)
+                status = gr.HTML(visible=False)
 
         def _update_btn(upload, mic):
             return gr.update(interactive=(upload is not None or mic is not None))
@@ -555,13 +556,23 @@ def build_demo():
         def _generate(upload, mic):
             audio = upload if upload is not None else mic
             imgs, msg = generate(audio, DEFAULT_NUM_DISPLAY, DEFAULT_GUIDANCE_SCALE, DEFAULT_NUM_STEPS, DEFAULT_BASE_SEED)
-            visible = bool(msg)
-            return imgs, gr.update(value=msg, visible=visible)
+            if msg:
+                error_html = f'<div style="background:#fee2e2;border:1px solid #f87171;border-radius:8px;padding:12px 16px;color:#b91c1c;font-size:0.95em;">⚠️ {msg}</div>'
+                return imgs, gr.update(value=error_html, visible=True)
+            return imgs, gr.update(value="", visible=False)
 
         generate_btn.click(
+            fn=lambda u, m: (gr.update(value="Generating...", interactive=False, variant="secondary"), gr.update(value="", visible=False)),
+            inputs=[audio_upload, audio_mic],
+            outputs=[generate_btn, status],
+        ).then(
             fn=_generate,
             inputs=[audio_upload, audio_mic],
             outputs=[gallery, status],
+        ).then(
+            fn=lambda u, m: gr.update(value="Generate", interactive=(u is not None or m is not None), variant="primary"),
+            inputs=[audio_upload, audio_mic],
+            outputs=generate_btn,
         )
 
     return demo