Add macOS microphone selection and paste fixes

src-tauri/Info.plist

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+  <key>NSMicrophoneUsageDescription</key>
+  <string>VibeToText needs microphone access to capture speech for local dictation.</string>
+</dict>
+</plist>

src-tauri/src/audio.rs

@@ -1,6 +1,7 @@
 use anyhow::{anyhow, Result};
 use cpal::traits::{DeviceTrait, HostTrait, StreamTrait};
 use crossbeam_channel::{bounded, Receiver, Sender};
+use serde::Serialize;
 use std::sync::{
     atomic::{AtomicBool, Ordering},
     Arc,
@@ -16,6 +17,13 @@ pub struct Frame(pub Vec<f32>);
 #[derive(Clone, Copy, Debug)]
 pub struct Level(pub f32);
 
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct InputDeviceInfo {
+    pub name: String,
+    pub is_default: bool,
+}
+
 pub struct AudioCapture {
     stop: Arc<AtomicBool>,
     pub frames: Receiver<Frame>,
@@ -75,6 +83,27 @@ impl AudioCapture {
     }
 }
 
+pub fn input_devices() -> Result<Vec<InputDeviceInfo>> {
+    let host = cpal::default_host();
+    let default_name = host.default_input_device().and_then(|d| d.name().ok());
+    let mut devices = Vec::new();
+
+    for device in host.input_devices()? {
+        let Ok(name) = device.name() else {
+            continue;
+        };
+        let is_default = default_name.as_ref() == Some(&name);
+        devices.push(InputDeviceInfo { name, is_default });
+    }
+
+    devices.sort_by(|a, b| match (a.is_default, b.is_default) {
+        (true, false) => std::cmp::Ordering::Less,
+        (false, true) => std::cmp::Ordering::Greater,
+        _ => a.name.to_lowercase().cmp(&b.name.to_lowercase()),
+    });
+    Ok(devices)
+}
+
 fn run_stream(
     device: cpal::Device,
     config: cpal::SupportedStreamConfig,

src-tauri/src/inject.rs

@@ -1,13 +1,15 @@
 use anyhow::Result;
-use enigo::{Direction, Enigo, Key, Keyboard, Settings};
+#[cfg(not(target_os = "macos"))]
+use enigo::Key;
+use enigo::{Direction, Enigo, Keyboard, Settings};
 use tauri::AppHandle;
 use tauri_plugin_clipboard_manager::ClipboardExt;
 
 /// Type a string at the current cursor as if the user typed it.
 /// Used in stream mode for incremental partials.
 #[allow(dead_code)]
 pub fn type_text(s: &str) -> Result<()> {
-    let mut enigo = Enigo::new(&Settings::default())?;
+    let mut enigo = new_enigo()?;
     enigo.text(s)?;
     Ok(())
 }
@@ -74,15 +76,7 @@ pub fn paste_text(app: &AppHandle, s: &str) -> Result<()> {
         );
     }
 
-    let mut enigo = Enigo::new(&Settings::default())?;
-    #[cfg(target_os = "macos")]
-    let mod_key = Key::Meta;
-    #[cfg(not(target_os = "macos"))]
-    let mod_key = Key::Control;
-
-    enigo.key(mod_key, Direction::Press)?;
-    enigo.key(Key::Unicode('v'), Direction::Click)?;
-    enigo.key(mod_key, Direction::Release)?;
+    send_paste_shortcut()?;
 
     // Restore the user's previous clipboard contents. CRITICAL: there's
     // no API for "the foreground app finished consuming this paste,"
@@ -136,3 +130,33 @@ pub fn paste_text(app: &AppHandle, s: &str) -> Result<()> {
     }
     Ok(())
 }
+
+#[cfg(target_os = "macos")]
+fn send_paste_shortcut() -> Result<()> {
+    let mut enigo = new_enigo()?;
+    // macOS keycode 55 = Command, 9 = V. Using raw keycodes avoids Enigo's
+    // layout lookup path, which must run on the main dispatch queue.
+    enigo.raw(55, Direction::Press)?;
+    enigo.raw(9, Direction::Click)?;
+    enigo.raw(55, Direction::Release)?;
+    Ok(())
+}
+
+#[cfg(not(target_os = "macos"))]
+fn send_paste_shortcut() -> Result<()> {
+    let mut enigo = new_enigo()?;
+    enigo.key(Key::Control, Direction::Press)?;
+    enigo.key(Key::Unicode('v'), Direction::Click)?;
+    enigo.key(Key::Control, Direction::Release)?;
+    Ok(())
+}
+
+fn new_enigo() -> Result<Enigo> {
+    // The app asks for Accessibility separately. During dictation, repeatedly
+    // opening the system prompt is noisy and can steal focus from the target app.
+    let settings = Settings {
+        open_prompt_to_get_permissions: false,
+        ..Settings::default()
+    };
+    Ok(Enigo::new(&settings)?)
+}

src-tauri/src/lib.rs

@@ -69,6 +69,11 @@ async fn get_config(state: tauri::State<'_, Arc<AppState>>) -> Result<AppConfig,
     Ok(state.config.lock().await.clone())
 }
 
+#[tauri::command]
+fn list_input_devices() -> Result<Vec<audio::InputDeviceInfo>, String> {
+    audio::input_devices().map_err(|e| e.to_string())
+}
+
 #[tauri::command]
 async fn save_config(
     app: tauri::AppHandle,
@@ -1176,6 +1181,7 @@ pub fn run() {
         })
         .invoke_handler(tauri::generate_handler![
             get_config,
+            list_input_devices,
             save_config,
             toggle_dictation,
             current_backend,

src/index.html

@@ -172,8 +172,11 @@ <h2>Behavior</h2>
           </label>
           <label>
             <span>Microphone</span>
-            <input id="mic_device" placeholder="(system default)" />
+            <select id="mic_device">
+              <option value="">(system default)</option>
+            </select>
           </label>
+          <p id="mic_device_hint" class="hint small">Microphones: checking…</p>
           <label class="toggle compact">
             <input type="checkbox" id="auto_start" />
             <span class="toggle-track"><span class="toggle-thumb"></span></span>

src/main.js

@@ -25,7 +25,6 @@ const fields = [
   "toggle_settings_hotkey",
   "dictate_hotkey",
   "stt_toggle_hotkey",
-  "mic_device",
 ];
 
 async function loadConfig() {
@@ -47,6 +46,7 @@ async function loadConfig() {
     const el = document.getElementById(k);
     if (el) el.value = cfg[toCamel(k)] ?? "";
   }
+  await refreshMicrophones(cfg.micDevice || "");
   document.getElementById("trailing_space").checked = !!cfg.trailingSpace;
   // STT master toggle. cfg.sttEnabled defaults to true if missing on
   // older configs that didn't have the field.
@@ -116,6 +116,55 @@ function toCamel(s) {
   return s.replace(/_([a-z])/g, (_, c) => c.toUpperCase());
 }
 
+async function refreshMicrophones(selected = val("mic_device")) {
+  const select = document.getElementById("mic_device");
+  const hint = document.getElementById("mic_device_hint");
+  if (!select) return;
+
+  try {
+    const devices = await invoke("list_input_devices");
+    select.innerHTML = "";
+
+    const defaultDevice = devices.find((d) => d.isDefault);
+    select.appendChild(
+      new Option(
+        defaultDevice
+          ? `(system default — ${defaultDevice.name})`
+          : "(system default)",
+        ""
+      )
+    );
+
+    for (const device of devices) {
+      const label = device.isDefault ? `${device.name} — default` : device.name;
+      select.appendChild(new Option(label, device.name));
+    }
+
+    if (selected && !devices.some((d) => d.name === selected)) {
+      select.appendChild(new Option(`${selected} — not currently available`, selected));
+    }
+
+    select.value = selected || "";
+    select.disabled = devices.length === 0;
+    if (hint) {
+      hint.textContent =
+        devices.length === 0
+          ? "No microphones found"
+          : `${devices.length} microphone${devices.length === 1 ? "" : "s"} available`;
+    }
+  } catch (e) {
+    console.error("failed to list microphones:", e);
+    select.innerHTML = "";
+    select.appendChild(new Option("(system default)", ""));
+    select.value = selected || "";
+    if (selected) {
+      select.appendChild(new Option(`${selected} — saved`, selected));
+      select.value = selected;
+    }
+    if (hint) hint.textContent = "Microphones unavailable";
+  }
+}
+
 const HOTKEY_DEFAULTS = {
   toggle_settings_hotkey: "Ctrl+Alt+V",
   dictate_hotkey: "Ctrl+Alt+D",
@@ -282,15 +331,8 @@ function wireAutoSave() {
   if (auto) auto.addEventListener("change", () => scheduleSave({ immediate: true }));
   const stt = document.getElementById("stt_enabled");
   if (stt) stt.addEventListener("change", () => scheduleSave({ immediate: true }));
-  // Free-text fields — debounce so typing isn't one save per keystroke.
-  ["mic_device"].forEach((id) => {
-    const el = document.getElementById(id);
-    if (!el) return;
-    el.addEventListener("input", () => scheduleSave());
-    // Commit immediately when focus leaves so a debounced save in flight
-    // doesn't get stranded.
-    el.addEventListener("blur", () => scheduleSave({ immediate: true }));
-  });
+  const mic = document.getElementById("mic_device");
+  if (mic) mic.addEventListener("change", () => scheduleSave({ immediate: true }));
   // Multi-line text fields (initial prompt, custom dictionary).
   // Same debounce-on-input + commit-on-blur pattern.
   ["whisper_initial_prompt", "custom_dictionary"].forEach((id) => {
@@ -1054,6 +1096,7 @@ listen("settings-shown", () => {
   // Force reflow so the next class-add re-runs the keyframes.
   void body.offsetWidth;
   body.classList.add("show-pop");
+  refreshMicrophones();
 });
 
 // Kept for forward-compatibility — currently we always select "whisper".