diff --git a/static/boot.js b/static/boot.js
index 1ff1087a..cb069162 100644
--- a/static/boot.js
+++ b/static/boot.js
@@ -1026,7 +1026,10 @@ window._hermesTtsSynth=function(id, text, opts){
   let _browserTtsKeepAlive=null;
   let _browserTtsWatchdog=null;
   let _browserTtsSuppressNextErrorRearm=false;
-  const SILENCE_MS=1800; // auto-send after 1.8s silence
+  // Configurable via localStorage keys (set from dev console or a future settings panel).
+  //   hermes-voice-silence-ms   — pause duration before auto-send (ms, default 1800)
+  //   hermes-voice-continuous   — keep mic open across natural pauses ("true"/"false", default false)
+  const SILENCE_MS=parseInt(localStorage.getItem('hermes-voice-silence-ms'))||1800;
 
   function _clearBrowserTtsRecovery(){
     if(_browserTtsKeepAlive){
@@ -1086,7 +1089,7 @@ window._hermesTtsSynth=function(id, text, opts){
     _setState('listening');
 
     _recognition=new SpeechRecognition();
-    _recognition.continuous=false;
+    _recognition.continuous=localStorage.getItem('hermes-voice-continuous')==='true';
     _recognition.interimResults=true;
     _recognition.lang=(typeof _locale!=='undefined'&&_locale._speech)||'en-US';
 
diff --git a/tests/test_issue4761_voice_mode_config.py b/tests/test_issue4761_voice_mode_config.py
new file mode 100644
index 00000000..0446b183
--- /dev/null
+++ b/tests/test_issue4761_voice_mode_config.py
@@ -0,0 +1,85 @@
+"""Tests for #4761 — configurable voice-mode silence timeout and continuous recognition.
+
+The voice-mode loop currently hardcodes:
+  - SILENCE_MS = 1800 (1.8s pause before auto-send)
+  - _recognition.continuous = false (mic closes after each utterance)
+
+This module pins the fix: both values are now configurable via localStorage keys
+(hermes-voice-silence-ms, hermes-voice-continuous) with sensible defaults.
+"""
+
+import re
+from pathlib import Path
+
+REPO = Path(__file__).resolve().parents[1]
+
+
+def _boot_src() -> str:
+    return (REPO / "static" / "boot.js").read_text(encoding="utf-8")
+
+
+class TestVoiceModeSilenceMsConfig:
+    """SILENCE_MS must read from localStorage with 1800 fallback."""
+
+    def test_silence_ms_reads_local_storage_with_fallback(self):
+        src = _boot_src()
+        m = re.search(
+            r"const\s+SILENCE_MS\s*=\s*parseInt\s*\(\s*localStorage\.getItem\s*\(\s*'hermes-voice-silence-ms'\s*\)\s*\)\s*\|\|\s*1800",
+            src,
+        )
+        assert m, (
+            "SILENCE_MS must be defined as `parseInt(localStorage.getItem('hermes-voice-silence-ms')) || 1800` "
+            "so users can tune it via dev console or a future settings toggle."
+        )
+
+    def test_silence_ms_used_in_timeout(self):
+        src = _boot_src()
+        assert "SILENCE_MS" in src, "SILENCE_MS must still be referenced in the timeout call."
+
+
+class TestVoiceModeContinuousConfig:
+    """_recognition.continuous must read from localStorage."""
+
+    def test_continuous_reads_local_storage(self):
+        src = _boot_src()
+        assert (
+            "_recognition.continuous=localStorage.getItem('hermes-voice-continuous')==='true'"
+            in src
+        ), (
+            "_recognition.continuous must read from localStorage key "
+            "'hermes-voice-continuous' with default false. "
+            "Without this, users with natural mid-sentence pauses get cut off."
+        )
+
+    def test_continuous_true_behavior(self):
+        """When hermes-voice-continuous is 'true', the recognition stays open
+        across pauses, so the silence timer is the sole arbiter of send timing."""
+        src = _boot_src()
+        # The continuous flag must not replace or disable the silence timer logic.
+        assert (
+            "_silenceTimer=setTimeout" in src
+        ), "The silence timer must still exist for continuous mode send decision."
+
+
+class TestBootJsVoiceSectionIntegrity:
+    """Smoke checks — the surrounding voice-mode infrastructure is intact."""
+
+    def test_voice_mode_declares_silence_ms(self):
+        src = _boot_src()
+        assert "SILENCE_MS" in src, "SILENCE_MS constant must exist in boot.js"
+
+    def test_voice_mode_declares_recognition(self):
+        src = _boot_src()
+        assert "_recognition=new SpeechRecognition()" in src
+
+    def test_voice_mode_state_machine_present(self):
+        src = _boot_src()
+        for state in ("idle", "listening", "thinking", "speaking"):
+            assert f"'{state}'" in src, f"Voice mode state '{state}' must be referenced."
+
+    def test_voice_mode_patches_auto_read(self):
+        src = _boot_src()
+        assert "autoReadLastAssistant" in src, (
+            "voice mode must still override autoReadLastAssistant to pipe "
+            "response completion into _speakResponse."
+        )