fix: fix mic settings, fix calibration

Author: DavidHruby1

src/audio.py

@@ -18,6 +18,7 @@
 except OSError:
     sd = None  # type: ignore[assignment]
 
+from src.config import DEFAULT_RMS_THRESHOLD
 from src.utils import AppError, ScreamerError
 
 log = logging.getLogger(__name__)
@@ -54,6 +55,16 @@ def list_devices() -> list[AudioDevice]:
     return result
 
 
+def default_input_device_id() -> int | None:
+    """Return PortAudio's default input device ID, if one is configured."""
+    _require_sd()
+    default = sd.default.device
+    device_id = default[0] if isinstance(default, (list, tuple)) else default
+    if device_id is None or int(device_id) < 0:
+        return None
+    return int(device_id)
+
+
 class AudioRecorder:
     def __init__(self, device_id: int | None = None, sample_rate: int = SAMPLE_RATE) -> None:
         self._device_id = device_id
@@ -62,7 +73,7 @@ def __init__(self, device_id: int | None = None, sample_rate: int = SAMPLE_RATE)
         self._stream: Any = None
         self._lock = threading.Lock()
         self._start_time: float = 0.0
-        self._rms_threshold: float = 50.0
+        self._rms_threshold: float = DEFAULT_RMS_THRESHOLD
 
     @property
     def rms_threshold(self) -> float:
@@ -78,7 +89,7 @@ def is_recording(self) -> bool:
         return self._stream is not None
 
     def calibrate(self, duration: float = 2.0) -> float:
-        """Record ambient noise, return noise_floor * 2.0. Fallback: 50."""
+        """Record ambient noise and return a usable silence-gate threshold."""
         _require_sd()
         try:
             log.info("Calibrating RMS threshold for %.1fs...", duration)
@@ -92,15 +103,15 @@ def calibrate(self, duration: float = 2.0) -> float:
             sd.wait()
             noise_floor = float(np.sqrt(np.mean(recording.astype(np.float64) ** 2)))
             threshold = noise_floor * 2.0
-            if threshold < 1.0:
-                threshold = 50.0
+            if threshold < DEFAULT_RMS_THRESHOLD:
+                threshold = DEFAULT_RMS_THRESHOLD
             self._rms_threshold = threshold
             log.info("Calibration done: noise_floor=%.1f, threshold=%.1f", noise_floor, threshold)
             return threshold
         except Exception as e:
-            log.warning("Calibration failed: %s; using fallback 50", e)
-            self._rms_threshold = 50.0
-            return 50.0
+            log.warning("Calibration failed: %s; using fallback %.1f", e, DEFAULT_RMS_THRESHOLD)
+            self._rms_threshold = DEFAULT_RMS_THRESHOLD
+            return DEFAULT_RMS_THRESHOLD
 
     def _callback(self, indata: np.ndarray, frames: int, time_info, status) -> None:  # type: ignore[no-untyped-def]
         if status:

src/config.py

@@ -19,6 +19,8 @@
     "Return only the corrected text with no explanations."
 )
 
+DEFAULT_RMS_THRESHOLD = 5.0
+
 MOD_CONTROL = 0x0002
 MOD_ALT = 0x0001
 MOD_SHIFT = 0x0004
@@ -100,7 +102,7 @@ class AppConfig:
     post_type_key: str = "none"  # "none" | "enter" | "tab" | "space" | "backspace"
     audio_device_id: int | None = None
     audio_device_name: str = ""
-    rms_threshold: float = 50.0
+    rms_threshold: float = DEFAULT_RMS_THRESHOLD
     # STT primary
     stt_api_key: str = ""
     stt_base_url: str = ""

src/main.py

@@ -15,7 +15,7 @@
 from PySide6.QtGui import QIcon
 from PySide6.QtWidgets import QApplication, QMenu, QSystemTrayIcon
 
-from src.audio import AudioRecorder, resolve_device, list_devices
+from src.audio import AudioRecorder, default_input_device_id, list_devices, resolve_device
 from src.config import (
     HOTKEY_OPTIONS,
     POST_KEY_OPTIONS,
@@ -132,7 +132,14 @@ def __init__(self) -> None:
     def _get_device_list(self) -> list[tuple[int, str]]:
         """Return list of (device_id, name) for the settings dialog."""
         try:
-            return [(d.id, d.name) for d in list_devices()]
+            default_id = default_input_device_id()
+            devices = []
+            for d in list_devices():
+                name = d.name
+                if d.id == default_id:
+                    name = f"{name} (Default input)"
+                devices.append((d.id, name))
+            return devices
         except Exception as e:
             log.warning("Could not enumerate audio devices: %s", e)
             return []

src/settings_dialog.py

@@ -25,6 +25,7 @@
     QPlainTextEdit,
     QPushButton,
     QRadioButton,
+    QDoubleSpinBox,
     QTabWidget,
     QVBoxLayout,
     QWidget,
@@ -258,6 +259,14 @@ def _build_audio_tab(self) -> None:
         tab = QWidget()
         form = QFormLayout(tab)
 
+        hint = QLabel(
+            "No external mic needed: use System Default for the built-in laptop mic, "
+            "or pick the device named Microphone Array, Internal Mic, Realtek, DMIC, "
+            "or Intel Smart Sound. Avoid Monitor, Stereo Mix, HDMI, and output devices."
+        )
+        hint.setWordWrap(True)
+        form.addRow(hint)
+
         self._device_combo = QComboBox()
         self._populate_devices()
         form.addRow("Input Device:", self._device_combo)
@@ -268,6 +277,13 @@ def _build_audio_tab(self) -> None:
             self._calibrate_btn.setEnabled(False)
         form.addRow(self._calibrate_btn)
 
+        self._rms_spin = QDoubleSpinBox()
+        self._rms_spin.setRange(0.0, 32767.0)
+        self._rms_spin.setDecimals(1)
+        self._rms_spin.setSingleStep(1.0)
+        self._rms_spin.setSpecialValueText("Disabled")
+        form.addRow("RMS Threshold:", self._rms_spin)
+
         self._rms_label = QLabel("Threshold: —")
         form.addRow(self._rms_label)
 
@@ -313,6 +329,7 @@ def _populate(self, cfg: AppConfig) -> None:
         self._llm_fb_headers.setText(cfg.llm_fallback_custom_headers)
 
         # Audio
+        self._rms_spin.setValue(cfg.rms_threshold)
         self._rms_label.setText(f"Threshold: {cfg.rms_threshold:.1f}")
 
     def _collect(self) -> None:
@@ -351,6 +368,7 @@ def _collect(self) -> None:
 
         # Audio
         cfg.audio_device_id = self._device_combo.currentData()
+        cfg.rms_threshold = self._rms_spin.value()
         cfg.audio_device_name = ""
         if cfg.audio_device_id is not None:
             text = self._device_combo.currentText()
@@ -362,7 +380,7 @@ def _collect(self) -> None:
 
     def _populate_devices(self) -> None:
         """Fill the device combo from the pre-fetched device list."""
-        self._device_combo.addItem("(System Default)", None)
+        self._device_combo.addItem("System Default (usually built-in laptop mic)", None)
         for dev_id, dev_name in self._devices:
             self._device_combo.addItem(f"[{dev_id}] {dev_name}", dev_id)
 
@@ -387,6 +405,7 @@ def _on_calibrate(self) -> None:
             )
             threshold = self._calibrate_fn(device_id)
             self._working.rms_threshold = threshold
+            self._rms_spin.setValue(threshold)
             self._rms_label.setText(f"Threshold: {threshold:.1f}")
         except Exception as e:
             QMessageBox.warning(self, "Calibration Failed", str(e))
@@ -398,13 +417,9 @@ def _on_calibrate(self) -> None:
     def _set_dynamic_group_visible(self, group: QWidget, visible: bool) -> None:
         group.setVisible(visible)
 
-        parent = group.parentWidget()
-        if parent is not None and parent.layout() is not None:
-            parent.layout().invalidate()
-
         if self.layout() is not None:
-            self.layout().invalidate()
-        self.adjustSize()
+            self.layout().activate()
+        self.resize(self.sizeHint())
 
     # ------------------------------------------------------------------
     # Validation

tests/test_audio.py

@@ -0,0 +1,51 @@
+import unittest
+from unittest.mock import patch
+
+import numpy as np
+
+from src.audio import AudioRecorder
+from src.config import DEFAULT_RMS_THRESHOLD
+
+
+class FakeSoundDevice:
+    def __init__(self, recording: np.ndarray | None = None, error: Exception | None = None) -> None:
+        self.recording = recording
+        self.error = error
+
+    def rec(self, *args, **kwargs):
+        if self.error is not None:
+            raise self.error
+        return self.recording
+
+    def wait(self) -> None:
+        return None
+
+
+class AudioCalibrationTests(unittest.TestCase):
+    def test_calibration_uses_low_default_for_silent_input(self) -> None:
+        fake_sd = FakeSoundDevice(np.zeros((16000, 1), dtype=np.int16))
+
+        with patch("src.audio.sd", fake_sd):
+            threshold = AudioRecorder().calibrate(1.0)
+
+        self.assertEqual(threshold, DEFAULT_RMS_THRESHOLD)
+
+    def test_calibration_scales_audible_noise_floor(self) -> None:
+        fake_sd = FakeSoundDevice(np.full((16000, 1), 10, dtype=np.int16))
+
+        with patch("src.audio.sd", fake_sd):
+            threshold = AudioRecorder().calibrate(1.0)
+
+        self.assertEqual(threshold, 20.0)
+
+    def test_calibration_failure_uses_low_default(self) -> None:
+        fake_sd = FakeSoundDevice(error=RuntimeError("no mic"))
+
+        with patch("src.audio.sd", fake_sd):
+            threshold = AudioRecorder().calibrate(1.0)
+
+        self.assertEqual(threshold, DEFAULT_RMS_THRESHOLD)
+
+
+if __name__ == "__main__":
+    unittest.main()