diff --git a/src/fenrirscreenreader/speechDriver/hardwareSerialDriver.py b/src/fenrirscreenreader/speechDriver/hardwareSerialDriver.py
index 12c66d90..8eb05b89 100644
--- a/src/fenrirscreenreader/speechDriver/hardwareSerialDriver.py
+++ b/src/fenrirscreenreader/speechDriver/hardwareSerialDriver.py
@@ -322,31 +322,42 @@ class hardware_serial_driver(speech_driver):
             )
             return [device]
         devices = []
+        seen_devices = set()
         for pattern in (
             "/dev/serial/by-id/*",
             "/dev/ttyACM*",
             "/dev/ttyUSB*",
+            "/dev/ttyS0",
+            "/dev/ttyS1",
         ):
             matches = sorted(glob.glob(pattern))
+            unique_matches = []
+            for match in matches:
+                real_device = os.path.realpath(match)
+                if real_device in seen_devices:
+                    continue
+                seen_devices.add(real_device)
+                unique_matches.append(match)
             self._debug(
                 f"Hardware speech auto scan {pattern}: {matches}",
                 debug.DebugLevel.INFO,
                 on_any_level=True,
             )
-            if matches:
-                if len(matches) > 1:
+            if unique_matches:
+                if len(unique_matches) > 1:
                     self._debug(
                         "Hardware speech auto found multiple devices for "
-                        f"{pattern}: {matches}; probing in order",
+                        f"{pattern}: {unique_matches}; probing in order",
                         debug.DebugLevel.WARNING,
                         on_any_level=True,
                     )
                 self._debug(
-                    f"Hardware speech auto candidate devices: {matches}",
+                    "Hardware speech auto candidate devices: "
+                    f"{unique_matches}",
                     debug.DebugLevel.INFO,
                     on_any_level=True,
                 )
-                devices.extend(matches)
+                devices.extend(unique_matches)
         return devices
 
     def _termios_baud_rate(self, baud_rate):
diff --git a/src/fenrirscreenreader/speechDriver/litetalkDriver.py b/src/fenrirscreenreader/speechDriver/litetalkDriver.py
index 75397781..cd6def28 100644
--- a/src/fenrirscreenreader/speechDriver/litetalkDriver.py
+++ b/src/fenrirscreenreader/speechDriver/litetalkDriver.py
@@ -12,6 +12,7 @@ from fenrirscreenreader.speechDriver.hardwareSerialDriver import (
 class driver(hardware_serial_driver):
     cancel_command = b"\x18"
     hardware_probe_command = b"\x01I"
+    hardware_probe_timeout = 3.5
 
     def _speak_bytes(self, text):
         return self._clean_text(text).encode("ascii", errors="replace") + b"\r"
diff --git a/tests/unit/test_hardware_speech_drivers.py b/tests/unit/test_hardware_speech_drivers.py
index 0e7c36c5..fdfd9795 100644
--- a/tests/unit/test_hardware_speech_drivers.py
+++ b/tests/unit/test_hardware_speech_drivers.py
@@ -166,6 +166,9 @@ def test_auto_device_detection_prefers_probe_response(monkeypatch):
 
     def fake_select(readable, writable, exceptional, timeout):
         port = readable[0]
+        if timeout == 0:
+            return [], writable, exceptional
+        assert timeout == 3.5
         if port == 101 and writes.get(port) == b"\x01I":
             return readable, writable, exceptional
         return [], writable, exceptional
@@ -273,6 +276,70 @@ def test_auto_device_detection_fails_without_probe_response(
     assert speech_driver.device == "auto"
 
 
+def test_auto_device_detection_deduplicates_serial_by_id(monkeypatch):
+    opened_devices = []
+
+    def fake_glob(pattern):
+        if pattern == "/dev/serial/by-id/*":
+            return ["/dev/serial/by-id/modem0"]
+        if pattern == "/dev/ttyUSB*":
+            return ["/dev/ttyUSB0"]
+        return []
+
+    def fake_realpath(path):
+        if path in ("/dev/serial/by-id/modem0", "/dev/ttyUSB0"):
+            return "/dev/ttyUSB0"
+        return path
+
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.glob.glob",
+        fake_glob,
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.os.path.realpath",
+        fake_realpath,
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.os.open",
+        lambda device, flags: opened_devices.append(device)
+        or len(opened_devices),
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.os.close",
+        lambda port: None,
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.os.write",
+        lambda port, data: len(data),
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.select.select",
+        lambda readable, writable, exceptional, timeout: (
+            [],
+            writable,
+            exceptional,
+        ),
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.termios.tcgetattr",
+        lambda port: [0, 0, 0, 0, 0, 0, [0] * 32],
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.termios.tcsetattr",
+        lambda port, when, attrs: None,
+    )
+    monkeypatch.setattr(
+        "fenrirscreenreader.speechDriver.hardwareSerialDriver.tty.setraw",
+        lambda port: None,
+    )
+
+    speech_driver = litetalkDriver.driver()
+    with pytest.raises(RuntimeError, match="hardware speech device"):
+        speech_driver.initialize(build_environment("auto"))
+
+    assert opened_devices == ["/dev/serial/by-id/modem0"]
+
+
 def test_auto_device_detection_skips_termios_failures(monkeypatch):
     opened_ports = []
     closed_ports = []