Feat/multilingual whisper

Package.swift

@@ -17,14 +17,20 @@ let package = Package(
         .executableTarget(
             name: "MiniWhisper",
             dependencies: [
-                "FluidAudio"
+                "FluidAudio",
+                "whisper"
             ],
             path: "Sources/MiniWhisper",
             exclude: ["Resources"],
             swiftSettings: [
                 .enableExperimentalFeature("StrictConcurrency")
             ]
         ),
+        .binaryTarget(
+            name: "whisper",
+            url: "https://github.com/andyhtran/MiniWhisper/releases/download/whisper-xcframework-1.0/whisper.xcframework.zip",
+            checksum: "866b43e4a3f31d1f898c7300d36e786841723e7be5a0fcdaa5879daea2f4389d"
+        ),
         .testTarget(
             name: "MiniWhisperTests",
             dependencies: ["MiniWhisper"],

README.md

@@ -1,6 +1,6 @@
 # MiniWhisper
 
-A minimal macOS menu bar app for voice-to-text using the [Parakeet](https://github.com/FluidInference/FluidAudio) model. Press a hotkey, speak, and the transcription is automatically pasted into the active app.
+A minimal macOS menu bar app for voice-to-text with fast English transcription via [Parakeet](https://github.com/FluidInference/FluidAudio) and multilingual transcription via [whisper.cpp](https://github.com/ggml-org/whisper.cpp). Press a hotkey, speak, and the transcription is automatically pasted into the active app.
 
 ![macOS 14.0+](https://img.shields.io/badge/macOS-14.0%2B-blue)
 ![Swift 6.0+](https://img.shields.io/badge/Swift-6.0%2B-orange)
@@ -51,7 +51,8 @@ just dev
 - **Text replacements** — auto-correct words or phrases after transcription
 - **Recording history** — browse and copy recent transcriptions
 - **Usage stats** — track recordings, speaking time, word count, and average WPM
-- **On-device** — all processing happens locally via the Parakeet model, nothing leaves your Mac
+- **Multiple models** — switch between fast English-only (Parakeet) and multilingual auto-detect (whisper.cpp)
+- **On-device** — all processing happens locally on your Mac, nothing leaves your device
 
 ## Build commands
 

Scripts/build-app.sh

@@ -30,6 +30,22 @@ mkdir -p "$APP_BUNDLE/Contents/Resources"
 
 cp "$BUILD_DIR/$APP_NAME" "$APP_BUNDLE/Contents/MacOS/"
 
+# Embed whisper.framework with proper macOS versioned structure
+FRAMEWORKS_DIR="$APP_BUNDLE/Contents/Frameworks"
+if [ -d "$BUILD_DIR/whisper.framework" ]; then
+    mkdir -p "$FRAMEWORKS_DIR/whisper.framework/Versions/A"
+    cp -R "$BUILD_DIR/whisper.framework/Versions/A/Headers" "$FRAMEWORKS_DIR/whisper.framework/Versions/A/"
+    cp -R "$BUILD_DIR/whisper.framework/Versions/A/Modules" "$FRAMEWORKS_DIR/whisper.framework/Versions/A/"
+    cp -R "$BUILD_DIR/whisper.framework/Versions/A/Resources" "$FRAMEWORKS_DIR/whisper.framework/Versions/A/"
+    cp "$BUILD_DIR/whisper.framework/Versions/A/whisper" "$FRAMEWORKS_DIR/whisper.framework/Versions/A/"
+    ln -sf A "$FRAMEWORKS_DIR/whisper.framework/Versions/Current"
+    ln -sf Versions/Current/Headers "$FRAMEWORKS_DIR/whisper.framework/Headers"
+    ln -sf Versions/Current/Modules "$FRAMEWORKS_DIR/whisper.framework/Modules"
+    ln -sf Versions/Current/Resources "$FRAMEWORKS_DIR/whisper.framework/Resources"
+    ln -sf Versions/Current/whisper "$FRAMEWORKS_DIR/whisper.framework/whisper"
+    install_name_tool -add_rpath "@executable_path/../Frameworks" "$APP_BUNDLE/Contents/MacOS/$APP_NAME" 2>/dev/null || true
+fi
+
 cat > "$APP_BUNDLE/Contents/Info.plist" << PLIST
 <?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">

Scripts/sign-and-notarize.sh

@@ -12,6 +12,11 @@ SIGNING_ID="${CODESIGN_IDENTITY:?Set CODESIGN_IDENTITY to your Developer ID Appl
 echo "==> Building release..."
 bash "$ROOT/Scripts/build-app.sh" release
 
+echo "==> Signing embedded frameworks..."
+for fw in "$APP_BUNDLE"/Contents/Frameworks/*.framework; do
+    [ -d "$fw" ] && codesign --force --timestamp --options runtime --sign "$SIGNING_ID" "$fw"
+done
+
 echo "==> Signing with: $SIGNING_ID"
 codesign --force --timestamp --options runtime \
     --sign "$SIGNING_ID" \

Sources/MiniWhisper/AppState.swift

@@ -8,13 +8,15 @@ import UserNotifications
 final class AppState: Sendable {
     let recorder = AudioRecorder()
     let parakeet = ParakeetProvider()
+    let whisper = WhisperProvider()
     let recordingStore = RecordingStore()
     let analyticsStore = AnalyticsStore()
     let permissions = PermissionsManager()
     let pasteboard = PasteboardService()
     let toast = ToastWindowController.shared
 
     var replacementSettings = ReplacementSettings.load()
+    var transcriptionMode: TranscriptionMode = TranscriptionModeStorage.load()
 
     let maxRecordingDuration: TimeInterval = 600.0  // 10 minutes
     var warningDuration: TimeInterval { maxRecordingDuration * 0.8 }  // 8 minutes
@@ -23,18 +25,65 @@ final class AppState: Sendable {
     private var durationCheckTimer: Timer?
     private var currentRecordingId: String?
 
-    // Callbacks for HotkeyManager to track recording state
     var onRecordingStarted: (() -> Void)?
     var onRecordingEnded: (() -> Void)?
 
-    var isModelLoaded: Bool { parakeet.isInitialized }
+    var isModelLoaded: Bool {
+        switch transcriptionMode {
+        case .english: return parakeet.isInitialized
+        case .multilingual: return whisper.isInitialized
+        }
+    }
+
+    var isModelDownloading: Bool { whisper.isDownloading }
+    var modelDownloadProgress: Double { whisper.downloadProgress }
 
     // MARK: - Initialization
 
     func preloadModel() {
         Task {
             do {
-                try await parakeet.initialize()
+                switch transcriptionMode {
+                case .english:
+                    try await parakeet.initialize()
+                case .multilingual:
+                    guard whisper.modelExists else { return }
+                    try await whisper.initialize()
+                }
+            } catch {
+                toast.showError(title: "Model Load Failed", message: error.localizedDescription)
+            }
+        }
+    }
+
+    func switchTranscriptionMode(to mode: TranscriptionMode) {
+        guard mode != transcriptionMode else { return }
+
+        if recorder.state.isRecording {
+            toast.showError(title: "Cannot Switch", message: "Stop recording before switching models.")
+            return
+        }
+
+        if recorder.state == .processing {
+            return
+        }
+
+        switch transcriptionMode {
+        case .english: parakeet.unload()
+        case .multilingual: whisper.unload()
+        }
+
+        transcriptionMode = mode
+        TranscriptionModeStorage.save(mode)
+
+        Task {
+            do {
+                switch mode {
+                case .english:
+                    try await parakeet.initialize()
+                case .multilingual:
+                    try await whisper.initialize()
+                }
             } catch {
                 toast.showError(title: "Model Load Failed", message: error.localizedDescription)
             }
@@ -53,7 +102,7 @@ final class AppState: Sendable {
 
     func startRecording() {
         guard recorder.state.isIdle else { return }
-        guard parakeet.isInitialized else {
+        guard isModelLoaded else {
             toast.showError(title: "Model Not Ready", message: "Please wait for the model to finish loading.")
             return
         }
@@ -113,7 +162,13 @@ final class AppState: Sendable {
 
     private func transcribe(audioURL: URL, recordingId: String, duration: TimeInterval, sampleRate: Double) async {
         do {
-            let result = try await parakeet.transcribe(audioURL: audioURL)
+            let result: TranscriptionResult
+            switch transcriptionMode {
+            case .english:
+                result = try await parakeet.transcribe(audioURL: audioURL)
+            case .multilingual:
+                result = try await whisper.transcribe(audioURL: audioURL)
+            }
 
             // Guard against stale callback: if the user rapid-tapped and started a new
             // recording while transcription was in-flight, the state has moved on.
@@ -156,8 +211,8 @@ final class AppState: Sendable {
                     transcriptionDuration: result.duration
                 ),
                 configuration: RecordingConfiguration(
-                    voiceModel: "Parakeet",
-                    language: "en"
+                    voiceModel: result.model,
+                    language: result.language
                 )
             )
 
@@ -186,7 +241,7 @@ final class AppState: Sendable {
                 ),
                 transcription: nil,
                 configuration: RecordingConfiguration(
-                    voiceModel: "Parakeet",
+                    voiceModel: transcriptionMode == .english ? "Parakeet" : "Whisper",
                     language: "en"
                 )
             )

Sources/MiniWhisper/Models/CustomShortcut.swift

@@ -54,10 +54,10 @@ struct CustomShortcut: Codable, Equatable, Hashable {
 
     var compactDisplayString: String {
         var str = ""
-        if control { str += "⌃" }
-        if option { str += "⌥" }
-        if shift { str += "⇧" }
-        if command { str += "⌘" }
+        if control { str += "Ctrl+" }
+        if option { str += "Option+" }
+        if shift { str += "Shift+" }
+        if command { str += "Cmd+" }
         if fn { str += "Fn+" }
         str += keyCodeDisplayName
         return str
@@ -215,7 +215,7 @@ final class CustomShortcutStorage {
 
     static func defaultShortcuts() -> [CustomShortcutName: CustomShortcut] {
         [
-            .toggleRecording: CustomShortcut(keyCode: UInt16(kVK_ANSI_Grave), option: true),  // Option + `
+            .toggleRecording: CustomShortcut(keyCode: UInt16(kVK_ANSI_R), option: true, shift: true),
             .cancelRecording: CustomShortcut(keyCode: UInt16(kVK_Escape)),  // Escape
         ]
     }

Sources/MiniWhisper/Models/TranscriptionMode.swift

@@ -0,0 +1,22 @@
+import Foundation
+
+enum TranscriptionMode: String, Codable, Sendable {
+    case english
+    case multilingual
+}
+
+struct TranscriptionModeStorage: Sendable {
+    private static let storageKey = "TranscriptionMode"
+
+    static func load() -> TranscriptionMode {
+        guard let raw = UserDefaults.standard.string(forKey: storageKey),
+              let mode = TranscriptionMode(rawValue: raw) else {
+            return .english
+        }
+        return mode
+    }
+
+    static func save(_ mode: TranscriptionMode) {
+        UserDefaults.standard.set(mode.rawValue, forKey: storageKey)
+    }
+}

Sources/MiniWhisper/Services/ParakeetProvider.swift

@@ -9,6 +9,11 @@ final class ParakeetProvider {
 
     var isInitialized: Bool { asrManager != nil }
 
+    func unload() {
+        asrManager = nil
+        initializationTask = nil
+    }
+
     func initialize() async throws {
         if asrManager != nil { return }