Shofar

Cross-Platform Voice-to-Text

Speak naturally. Type instantly.

Features • Installation • Usage • Models • Architecture

---

    ┌─────────────────────────────────┐
    │       🎙 Recording...           │
    │    ▁▂▃▅▆▇█▇▆▅▃▂▁▂▃▅▆▇█▇▆▅     │
    │            0:03                 │
    └─────────────────────────────────┘
              ⬇️ Press hotkey
    ┌─────────────────────────────────┐
    │  📝 Result                      │
    │  ┌───────────────────────────┐  │
    │  │ Ваш распознанный текст    │  │
    │  │ появится здесь            │  │
    │  └───────────────────────────┘  │
    │     [Вставить]    [Копировать]  │
    └─────────────────────────────────┘

Press hotkey → Speak → Text appears in any application

---

💡 Why Shofar?

Most voice input solutions are either cloud-based (privacy concerns), require complex setup, or don't integrate with the desktop. Shofar is different:

Problem	Shofar Solution
🔐 Privacy concerns	100% Offline — voice never leaves your machine
📦 Complex dependencies	Single Binary — no Python, Docker, or npm
🖥️ Limited integration	System-wide — works in any app, any text field
⚡ Speed vs accuracy trade-off	Hot-swap Models — switch engines on the fly

---

✨ Features

🎯 Core Push-to-talk with customizable hotkey Real-time waveform visualization Auto-insert into active window System tray with status indication	🧠 Engines Whisper (OpenAI) — best accuracy Vosk — fastest, lowest latency LLM correction — fix recognition errors
🌍 Languages 🇷🇺 Russian (primary) 🇬🇧 English 🔄 Auto-detection	🎨 Interface Dark theme floating window Editable results before insert Copy to clipboard option Desktop notifications

---

📦 Installation

Quick Install (Recommended)

curl -fsSL https://raw.githubusercontent.com/RooTooZ/shofar/master/install.sh | sh

Or download from Releases.

---

Build from Source

1. Install Dependencies

🐧 Arch Linux

sudo pacman -S portaudio libayatana-appindicator gtk3 pkg-config cmake xdotool

For Wayland:

sudo pacman -S wtype

🐧 Ubuntu / Debian

sudo apt install gcc pkg-config cmake xdotool \
    libportaudio2 portaudio19-dev \
    libayatana-appindicator3-dev libgtk-3-dev \
    libwayland-dev libx11-dev libx11-xcb-dev \
    libxkbcommon-x11-dev libgles2-mesa-dev \
    libegl1-mesa-dev libffi-dev libxcursor-dev libvulkan-dev

For Wayland:

sudo apt install wtype

🍎 macOS

brew install portaudio pkg-config cmake

🪟 Windows

Requires:

• MinGW-w64 or MSYS2
• CMake
• PortAudio

# Using MSYS2
pacman -S mingw-w64-x86_64-portaudio mingw-w64-x86_64-cmake

2. Build

git clone https://github.com/RooTooZ/shofar.git
cd shofar

# Build everything (whisper.cpp + llama.cpp + binary)
make all

# Install to ~/.local/bin
make install

3. Download a Model

Models download automatically via Settings UI, or:

make download-model-tiny    # 75 MB  — fast
make download-model-small   # 466 MB — balanced
make download-model-turbo   # 574 MB — best quality

---

🚀 Usage

shofar    # or ./bin/shofar

Quick Start

Step	Action
1	Press Ctrl+Shift+Space
2	🎙️ Speak while window is visible
3	Press hotkey again to stop
4	✏️ Edit text if needed
5	Enter = insert, Esc = cancel

Keyboard Shortcuts

Key	Action
Ctrl+Shift+Space	Start / Stop recording
Enter	Insert text into active window
Esc	Cancel and close

Tray Menu

Right-click tray icon for:

• ⚙️ Settings — models, hotkey, language
• 🔔 Notifications — toggle on/off
• ❌ Quit

---

🧠 Models

Speech Recognition

Engine	Model	Size	Speed	Accuracy	Best For
Whisper	tiny	75 MB	⚡⚡⚡	★★☆☆	Quick commands
Whisper	small	466 MB	⚡⚡	★★★☆	Daily use
Whisper	turbo	574 MB	⚡	★★★★	Documents
Vosk	small-ru	45 MB	⚡⚡⚡⚡	★★☆☆	Real-time

LLM Text Correction (Optional)

Model	Size	Description
Qwen2.5-1.5B	1.1 GB	Fixes punctuation & typos

---

🏗 Architecture

┌─────────────────────────────────────────────────────────────────┐
│                           SHOFAR                                │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│   ┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐ │
│   │  Hotkey  │ -> │  Audio   │ -> │  Speech  │ -> │   Input  │ │
│   │  Handler │    │ Recorder │    │  Engine  │    │ Emulator │ │
│   └──────────┘    └──────────┘    └──────────┘    └──────────┘ │
│        │                              │                         │
│        v                              v                         │
│   ┌──────────┐                  ┌──────────┐                    │
│   │   Tray   │                  │   LLM    │                    │
│   │   Icon   │                  │ Corrector│                    │
│   └──────────┘                  └──────────┘                    │
│                                                                 │
├─────────────────────────────────────────────────────────────────┤
│  whisper.cpp  │  llama.cpp  │  vosk-api  │  PortAudio  │  Gio  │
└─────────────────────────────────────────────────────────────────┘

Project Structure

shofar/
├── cmd/shofar/            # Entry point
├── internal/
│   ├── app/               # Main coordinator
│   ├── audio/             # PortAudio recording
│   ├── speech/            # Engine factory (Whisper/Vosk)
│   ├── llm/               # LLM text correction
│   ├── hotkey/            # Global hotkey
│   ├── tray/              # System tray
│   ├── waveform/          # Recording UI
│   ├── settings/          # Settings UI
│   ├── input/             # xdotool/wtype wrapper
│   └── i18n/              # Translations
└── third_party/           # whisper.cpp, llama.cpp, vosk

Tech Stack

Component	Technology
Language	Go 1.21+ (CGO)
Speech	whisper.cpp, Vosk
LLM	llama.cpp
Audio	PortAudio
GUI	Gio
Tray	systray
Text Input	xdotool (X11) / wtype (Wayland) / CGEventPost (macOS) / SendInput (Windows)

---

⚙️ Configuration

Config stored at ~/.local/bin/config.json:

{
  "model_id": "whisper-small",
  "language": "ru",
  "hotkey": { "key": "Space", "modifiers": ["Ctrl", "Shift"] },
  "llm_enabled": false,
  "notifications": true
}

---

🛠 Development

make whisper-lib    # Build whisper.cpp
make llama-lib      # Build llama.cpp
make build          # Build application
make run            # Run for testing
make clean          # Clean artifacts

---

🤝 Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push branch (git push origin feature/amazing)
5. Open Pull Request

---

📄 License

MIT License — use freely, attribution appreciated.

---

⬆ Back to top

Made with ❤️ for desktop users everywhere

Star ⭐ if you find it useful!