diff --git a/routes/listening_post.py b/routes/listening_post.py
index 17cdd22..cded266 100644
--- a/routes/listening_post.py
+++ b/routes/listening_post.py
@@ -665,25 +665,41 @@ def scanner_loop_power():
         logger.info("Power sweep scanner thread stopped")
 
 
-def _start_audio_stream(frequency: float, modulation: str):
-    """Start audio streaming at given frequency."""
-    global audio_process, audio_rtl_process, audio_running, audio_frequency, audio_modulation
-
-    with audio_lock:
-        # Stop any existing stream
-        _stop_audio_stream_internal()
+def _start_audio_stream(
+    frequency: float,
+    modulation: str,
+    *,
+    device: int | None = None,
+    sdr_type: str | None = None,
+    gain: int | None = None,
+    squelch: int | None = None,
+    bias_t: bool | None = None,
+):
+    """Start audio streaming at given frequency."""
+    global audio_process, audio_rtl_process, audio_running, audio_frequency, audio_modulation
+
+    with audio_lock:
+        # Stop any existing stream
+        _stop_audio_stream_internal()
 
         ffmpeg_path = find_ffmpeg()
         if not ffmpeg_path:
             logger.error("ffmpeg not found")
             return
 
-        # Determine SDR type and build appropriate command
-        sdr_type_str = scanner_config.get('sdr_type', 'rtlsdr')
-        try:
-            sdr_type = SDRType(sdr_type_str)
-        except ValueError:
-            sdr_type = SDRType.RTL_SDR
+        # Snapshot runtime tuning config so the spawned demod command cannot
+        # drift if shared scanner_config changes while startup is in-flight.
+        device_index = int(device if device is not None else scanner_config.get('device', 0))
+        gain_value = int(gain if gain is not None else scanner_config.get('gain', 40))
+        squelch_value = int(squelch if squelch is not None else scanner_config.get('squelch', 0))
+        bias_t_enabled = bool(scanner_config.get('bias_t', False) if bias_t is None else bias_t)
+        sdr_type_str = str(sdr_type if sdr_type is not None else scanner_config.get('sdr_type', 'rtlsdr')).lower()
+
+        # Determine SDR type and build appropriate command
+        try:
+            sdr_type = SDRType(sdr_type_str)
+        except ValueError:
+            sdr_type = SDRType.RTL_SDR
 
         # Set sample rates based on modulation
         if modulation == 'wfm':
@@ -707,41 +723,41 @@ def _start_audio_stream(frequency: float, modulation: str):
             freq_hz = int(frequency * 1e6)
             sdr_cmd = [
                 rtl_fm_path,
-                '-M', _rtl_fm_demod_mode(modulation),
-                '-f', str(freq_hz),
-                '-s', str(sample_rate),
-                '-r', str(resample_rate),
-                '-g', str(scanner_config['gain']),
-                '-d', str(scanner_config['device']),
-                '-l', str(scanner_config['squelch']),
-            ]
-            if scanner_config.get('bias_t', False):
-                sdr_cmd.append('-T')
-            # Omit explicit filename: rtl_fm defaults to stdout.
-            # (Some builds intermittently stall when '-' is passed explicitly.)
-        else:
-            # Use SDR abstraction layer for HackRF, Airspy, LimeSDR, SDRPlay
+                '-M', _rtl_fm_demod_mode(modulation),
+                '-f', str(freq_hz),
+                '-s', str(sample_rate),
+                '-r', str(resample_rate),
+                '-g', str(gain_value),
+                '-d', str(device_index),
+                '-l', str(squelch_value),
+            ]
+            if bias_t_enabled:
+                sdr_cmd.append('-T')
+            # Omit explicit filename: rtl_fm defaults to stdout.
+            # (Some builds intermittently stall when '-' is passed explicitly.)
+        else:
+            # Use SDR abstraction layer for HackRF, Airspy, LimeSDR, SDRPlay
             rx_fm_path = find_rx_fm()
             if not rx_fm_path:
-                logger.error(f"rx_fm not found - required for {sdr_type.value}. Install SoapySDR utilities.")
-                return
-
-            # Create device and get command builder
-            device = SDRFactory.create_default_device(sdr_type, index=scanner_config['device'])
-            builder = SDRFactory.get_builder(sdr_type)
-
-            # Build FM demod command
-            sdr_cmd = builder.build_fm_demod_command(
-                device=device,
-                frequency_mhz=frequency,
-                sample_rate=resample_rate,
-                gain=float(scanner_config['gain']),
-                modulation=modulation,
-                squelch=scanner_config['squelch'],
-                bias_t=scanner_config.get('bias_t', False)
-            )
-            # Ensure we use the found rx_fm path
-            sdr_cmd[0] = rx_fm_path
+                logger.error(f"rx_fm not found - required for {sdr_type.value}. Install SoapySDR utilities.")
+                return
+
+            # Create device and get command builder
+            sdr_device = SDRFactory.create_default_device(sdr_type, index=device_index)
+            builder = SDRFactory.get_builder(sdr_type)
+
+            # Build FM demod command
+            sdr_cmd = builder.build_fm_demod_command(
+                device=sdr_device,
+                frequency_mhz=frequency,
+                sample_rate=resample_rate,
+                gain=float(gain_value),
+                modulation=modulation,
+                squelch=squelch_value,
+                bias_t=bias_t_enabled,
+            )
+            # Ensure we use the found rx_fm path
+            sdr_cmd[0] = rx_fm_path
 
         encoder_cmd = [
             ffmpeg_path,
@@ -762,11 +778,11 @@ def _start_audio_stream(frequency: float, modulation: str):
         ]
 
         try:
-            # Use subprocess piping for reliable streaming.
-            # Log stderr to temp files for error diagnosis.
-            rtl_stderr_log = '/tmp/rtl_fm_stderr.log'
-            ffmpeg_stderr_log = '/tmp/ffmpeg_stderr.log'
-            logger.info(f"Starting audio: {frequency} MHz, mod={modulation}, device={scanner_config['device']}")
+            # Use subprocess piping for reliable streaming.
+            # Log stderr to temp files for error diagnosis.
+            rtl_stderr_log = '/tmp/rtl_fm_stderr.log'
+            ffmpeg_stderr_log = '/tmp/ffmpeg_stderr.log'
+            logger.info(f"Starting audio: {frequency} MHz, mod={modulation}, device={device_index}")
 
             # Retry loop for USB device contention (device may not be
             # released immediately after a previous process exits)
@@ -1432,7 +1448,15 @@ def start_audio() -> Response:
                 }), 409
             receiver_active_device = device
 
-        _start_audio_stream(frequency, modulation)
+        _start_audio_stream(
+            frequency,
+            modulation,
+            device=device,
+            sdr_type=sdr_type,
+            gain=gain,
+            squelch=squelch,
+            bias_t=bias_t,
+        )
 
         if audio_running:
             audio_source = 'process'