Nvidia parakeet cross platform

CHANGELOG.md

@@ -8,6 +8,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [1.2.16] - 2026-01-26
 
 ### Added
+- **NVIDIA Parakeet Support**: Fast local transcription via sherpa-onnx runtime with INT8 quantized models
+  - `parakeet-tdt-0.6b-v3`: Multilingual (25 languages), ~680MB
 - **Windows Push-to-Talk**: Native Windows key listener with low-level keyboard hook for true push-to-talk functionality
   - Supports compound hotkeys like `Ctrl+Shift+F11` or `CommandOrControl+Space`
   - Prebuilt binary automatically downloaded from GitHub releases
@@ -25,6 +27,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Dev Server Port Alignment**: Development server port configuration improved for consistency
 
 ### Fixed
+- **Whisper Model Path Resolution**: Fixed `large` and `turbo` model lookup failures by using registry-defined filenames (`ggml-large-v3.bin`, `ggml-large-v3-turbo.bin`) instead of hardcoded pattern
 - **Windows Production Build**: Fixed Windows production build issues with proper binary bundling
 - **Code Quality**: Various code quality improvements in download scripts and dev server management
 

CLAUDE.md

@@ -13,7 +13,7 @@ OpenWhispr is an Electron-based desktop dictation application that uses whisper.
 - **Desktop Framework**: Electron 36 with context isolation
 - **Database**: better-sqlite3 for local transcription history
 - **UI Components**: shadcn/ui with Radix primitives
-- **Speech Processing**: whisper.cpp (bundled native binary) + OpenAI API
+- **Speech Processing**: whisper.cpp + NVIDIA Parakeet (via sherpa-onnx) + OpenAI API
 - **Audio Processing**: FFmpeg (bundled via ffmpeg-static)
 
 ### Key Architectural Decisions
@@ -76,6 +76,8 @@ OpenWhispr is an Electron-based desktop dictation application that uses whisper.
 - **menuManager.js**: Application menu management
 - **tray.js**: System tray icon and menu
 - **whisper.js**: Local whisper.cpp integration and model management
+- **parakeet.js**: NVIDIA Parakeet model management via sherpa-onnx
+- **parakeetServer.js**: sherpa-onnx CLI wrapper for transcription
 - **windowConfig.js**: Centralized window configuration
 - **windowManager.js**: Window creation and lifecycle management
 
@@ -118,12 +120,28 @@ OpenWhispr is an Electron-based desktop dictation application that uses whisper.
   - GGML model downloads from HuggingFace
   - Models stored in `~/.cache/openwhispr/whisper-models/`
 
+### NVIDIA Parakeet Integration (via sherpa-onnx)
+
+- **parakeet.js**: Model management for NVIDIA Parakeet ASR models
+  - Uses sherpa-onnx runtime for cross-platform ONNX inference
+  - Bundled binaries in `resources/bin/sherpa-onnx-{platform}-{arch}`
+  - INT8 quantized models for efficient CPU inference
+  - Models stored in `~/.cache/openwhispr/parakeet-models/`
+  - Server pre-warming on startup when `LOCAL_TRANSCRIPTION_PROVIDER=nvidia` is set
+  - Provider preference persisted to `.env` via `saveAllKeysToEnvFile()` on server start/stop
+
+- **Available Models**:
+  - `parakeet-tdt-0.6b-v3`: Multilingual (25 languages), ~680MB
+
+- **Download URLs**: Models from sherpa-onnx ASR models release on GitHub
+
 ### Build Scripts (scripts/)
 
 - **download-whisper-cpp.js**: Downloads whisper.cpp binaries from GitHub releases
 - **download-llama-server.js**: Downloads llama.cpp server for local LLM inference
 - **download-nircmd.js**: Downloads nircmd.exe for Windows clipboard operations
 - **download-windows-key-listener.js**: Downloads prebuilt Windows key listener binary
+- **download-sherpa-onnx.js**: Downloads sherpa-onnx binaries for Parakeet support
 - **build-globe-listener.js**: Compiles macOS Globe key listener from Swift source
 - **build-windows-key-listener.js**: Compiles Windows key listener (for local development)
 - **run-electron.js**: Development script to launch Electron with proper environment
@@ -195,6 +213,10 @@ Settings stored in localStorage with these keys:
 - `hasCompletedOnboarding`: Onboarding completion flag
 - `customDictionary`: JSON array of words/phrases for improved transcription accuracy
 
+Environment variables persisted to `.env` (via `saveAllKeysToEnvFile()`):
+- `LOCAL_TRANSCRIPTION_PROVIDER`: Transcription engine (`nvidia` for Parakeet)
+- `PARAKEET_MODEL`: Selected Parakeet model name (e.g., `parakeet-tdt-0.6b-v3`)
+
 ### 6. Language Support
 
 58 languages supported (see src/utils/languages.ts):

README.md

@@ -27,6 +27,7 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 - 📱 **Control Panel**: Manage settings, view history, and configure API keys
 - 🗄️ **Transcription History**: SQLite database stores all your transcriptions locally
 - 🔧 **Model Management**: Download and manage local Whisper models (tiny, base, small, medium, large, turbo)
+- ⚡ **NVIDIA Parakeet**: Fast local transcription via sherpa-onnx (multilingual, 25 languages)
 - 🧹 **Model Cleanup**: One-click removal of cached Whisper models with uninstall hooks to keep disks tidy
 - 🌐 **Cross-Platform**: Works on macOS, Windows, and Linux
 - ⚡ **Automatic Pasting**: Transcribed text automatically pastes at your cursor location
@@ -380,7 +381,7 @@ open-whispr/
 - **Desktop**: Electron 36 with context isolation
 - **UI Components**: shadcn/ui with Radix primitives
 - **Database**: better-sqlite3 for local transcription storage
-- **Speech-to-Text**: OpenAI Whisper (powered by whisper.cpp for local, OpenAI API for cloud)
+- **Speech-to-Text**: OpenAI Whisper (whisper.cpp) + NVIDIA Parakeet (sherpa-onnx) for local, OpenAI API for cloud
 - **Icons**: Lucide React for consistent iconography
 
 ## Development
@@ -490,6 +491,21 @@ For local processing, OpenWhispr uses OpenAI's Whisper model via whisper.cpp - a
 
 **Upgrading from Python-based version**: If you previously used the Python-based Whisper, you'll need to re-download models in GGML format. You can safely delete the old Python environment (`~/.openwhispr/python/`) and PyTorch models (`~/.cache/whisper/`) to reclaim disk space.
 
+### Local Parakeet Setup (Alternative)
+
+OpenWhispr also supports NVIDIA Parakeet models via sherpa-onnx - a fast alternative to Whisper:
+
+1. **Bundled Binary**: sherpa-onnx is bundled with the app for all platforms
+2. **INT8 Quantized Models**: Efficient CPU inference
+3. **Models stored in**: `~/.cache/openwhispr/parakeet-models/`
+
+**Available Models**:
+- `parakeet-tdt-0.6b-v3`: Multilingual (25 languages), ~680MB
+
+**When to use Parakeet vs Whisper**:
+- **Parakeet**: Best for speed-critical use cases or lower-end hardware
+- **Whisper**: Best for quality-critical use cases or when you need specific model sizes
+
 ### Customization
 
 - **Hotkey**: Change in the Control Panel (default: backtick `) - fully customizable
@@ -602,6 +618,8 @@ OpenWhispr is actively maintained and ready for production use. Current version:
 
 - **[OpenAI Whisper](https://github.com/openai/whisper)** - The speech recognition model that powers both local and cloud transcription
 - **[whisper.cpp](https://github.com/ggerganov/whisper.cpp)** - High-performance C++ implementation of Whisper for local processing
+- **[NVIDIA Parakeet](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3)** - Fast ASR model for efficient local transcription
+- **[sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx)** - Cross-platform ONNX runtime for Parakeet model inference
 - **[Electron](https://www.electronjs.org/)** - Cross-platform desktop application framework
 - **[React](https://react.dev/)** - UI component library
 - **[shadcn/ui](https://ui.shadcn.com/)** - Beautiful UI components built on Radix primitives

electron-builder.json

@@ -88,7 +88,7 @@
     {
       "from": "resources/bin/",
       "to": "bin/",
-      "filter": ["whisper-cpp-*", "whisper-server-*", "llama-server-*", "windows-key-listener*", "*.dylib", "*.dll", "*.so*"]
+      "filter": ["whisper-cpp-*", "whisper-server-*", "llama-server-*", "sherpa-onnx-*", "windows-key-listener*", "*.dylib", "*.dll", "*.so*"]
     }
   ],
   "mac": {

main.js

@@ -44,6 +44,7 @@ const WindowManager = require("./src/helpers/windowManager");
 const DatabaseManager = require("./src/helpers/database");
 const ClipboardManager = require("./src/helpers/clipboard");
 const WhisperManager = require("./src/helpers/whisper");
+const ParakeetManager = require("./src/helpers/parakeet");
 const TrayManager = require("./src/helpers/tray");
 const IPCHandlers = require("./src/helpers/ipcHandlers");
 const UpdateManager = require("./src/updater");
@@ -58,6 +59,7 @@ let hotkeyManager = null;
 let databaseManager = null;
 let clipboardManager = null;
 let whisperManager = null;
+let parakeetManager = null;
 let trayManager = null;
 let updateManager = null;
 let globeKeyManager = null;
@@ -103,6 +105,7 @@ function initializeManagers() {
   databaseManager = new DatabaseManager();
   clipboardManager = new ClipboardManager();
   whisperManager = new WhisperManager();
+  parakeetManager = new ParakeetManager();
   trayManager = new TrayManager();
   updateManager = new UpdateManager();
   globeKeyManager = new GlobeKeyManager();
@@ -144,6 +147,7 @@ function initializeManagers() {
     databaseManager,
     clipboardManager,
     whisperManager,
+    parakeetManager,
     windowManager,
     updateManager,
     windowsKeyManager,
@@ -180,6 +184,19 @@ async function startApp() {
     debugLogger.debug("Whisper startup init error (non-fatal)", { error: err.message });
   });
 
+  // Initialize Parakeet manager at startup (don't await to avoid blocking)
+  // Settings can be provided via environment variables for server pre-warming:
+  // - LOCAL_TRANSCRIPTION_PROVIDER=nvidia to enable parakeet
+  // - PARAKEET_MODEL=parakeet-tdt-0.6b-v3 (model name)
+  const parakeetSettings = {
+    localTranscriptionProvider: process.env.LOCAL_TRANSCRIPTION_PROVIDER || "",
+    parakeetModel: process.env.PARAKEET_MODEL || "parakeet-tdt-0.6b-v3",
+  };
+  parakeetManager.initializeAtStartup(parakeetSettings).catch((err) => {
+    // Parakeet not being available at startup is not critical
+    debugLogger.debug("Parakeet startup init error (non-fatal)", { error: err.message });
+  });
+
   // Pre-warm llama-server if local reasoning is configured
   // Settings can be provided via environment variables:
   // - REASONING_PROVIDER=local to enable local reasoning
@@ -526,6 +543,10 @@ if (gotSingleInstanceLock) {
     if (whisperManager) {
       whisperManager.stopServer().catch(() => {});
     }
+    // Stop parakeet WS server if running
+    if (parakeetManager) {
+      parakeetManager.stopServer().catch(() => {});
+    }
     // Stop llama-server if running
     const modelManager = require("./src/helpers/modelManagerBridge").default;
     modelManager.stopServer().catch(() => {});