Robustly handle audio_ref dict or object in generate_audio endpoint

app.py

@@ -198,7 +198,14 @@ def generate_audio(
         raise gr.Error("Prompt is empty.")
     
     t0 = time.time()
-    ref_path = audio_ref.path if audio_ref and audio_ref.path and os.path.exists(audio_ref.path) else None
+    ref_path = None
+    if audio_ref:
+        if isinstance(audio_ref, dict):
+            ref_path = audio_ref.get("path")
+        elif hasattr(audio_ref, "path"):
+            ref_path = audio_ref.path
+    if ref_path and not os.path.exists(ref_path):
+        ref_path = None
     
     output = tempfile.mktemp(suffix=".wav", prefix="dramabox_")
     tts.generate_to_file(