..

app.py

@@ -198,11 +198,16 @@ def build_app() -> gr.Blocks:
                     sources    = ["microphone"],
                     streaming  = True,
                     type       = "numpy",
-                    label      = "Microphone (auto-detecting speech)",
+                    label      = "Microphone input",
                     interactive= True,
                     elem_id    = "audio-input",
                 )
 
+                gr.Markdown(
+                    "_Tip: click the microphone widget to grant browser permission, then speak. "
+                    "The call will start automatically on the first live audio input, or you can press 📞 Start Call._"
+                )
+
                 vad_dot = gr.Markdown("⚫ _mic idle_", elem_id="vad-dot")
 
                 with gr.Row():
@@ -338,6 +343,10 @@ def build_app() -> gr.Blocks:
             """
             state = _get_or_create_session(state)
 
+            if audio_chunk is not None and not state.call_active:
+                logger.info("Auto-starting call session on first live audio input.")
+                state.start_call()
+
             if not state.call_active:
                 # Return current state without processing
                 u = state._build_update()