SayKey

简体中文|English

SayKey is a tool that turns your speech into text. It works fast, accurately, and without the internet. It uses SenseVoice to do this.

---

Key Features

• ⚡ Super Fast: Convert speech to text in real-time.
• 🎯 Accurate: Enjoy precise transcriptions.
• 🔒 100% Offline: Your data stays on your device.
• ⌨️ Hotkey Activated: Start dictation with a simple keyboard shortcut.
• ✨ Smart Punctuation: Automatically adds punctuation to your text.
• 🛠️ Customizable: Easy-to-use settings for a personalized experience.

---

Quick Start Guide

1. Download SayKey
   Or visit the Releases page and download the latest SayKey.zip.
2. Extract the SayKey.zip file.
3. Run SayKey.exe.
4. If Windows Defender shows "Windows protected your PC":
   • Click More info → Run anyway.
5. Look for the white capsule-shaped icon above your taskbar.
6. Place the text cursor where you want to type.
7. Hold Ctrl+Q, speak, and release Ctrl+Q to convert speech to text!

For a detailed setup guide, check out the Installation Wiki.

---

System Requirements

• OS: Windows 10 or later
• Memory: At least 4 GB RAM (8 GB recommended)
• Disk: 1.5 GB free disk space
• CPU: x86-64 CPU with AVX support recommended

See the Installation Guide for the most up-to-date information.

---

Usage

Start voice typing

1. Make sure SayKey.exe is running and you see the capsule icon above the taskbar.
2. Focus any text input (Notepad, Word, browser, chat app, etc.).
3. Press and hold Ctrl+Q to start recording.
4. Speak clearly into your microphone.
5. Release Ctrl+Q. SayKey will recognize your speech and type the text at the cursor position.

Select microphone

1. Right-click the capsule icon.
2. Click Microphone.
3. Choose the device you want to use.

Change the hotkey

• Use the settings in the desktop app, or
• Call the HTTP API POST /set_hotkey if you are integrating SayKey programmatically.

---

For Developers

SayKey is open-source and we welcome contributions.

Clone the repository

git clone https://github.com/WenJing95/SayKey.git
cd SayKey

Dependencies

• Python 3.10
• Node.js (latest LTS)
• Git

Backend setup

cd backend
cd CT-Transformer-punctuation
pip install -e .
pip install -r requirements.txt

Start the backend server:

python main.py --sense-voice=./sherpa-onnx/model.int8.onnx --tokens=./sherpa-onnx/tokens.txt

You should see output similar to:

SayKey is running. Hold ctrl+q to start recording, release to recognize.
Important: Ensure the cursor is in the desired input location before using voice typing.
INFO:     Uvicorn running on http://localhost:58652 (Press CTRL+C to quit)

Command-line arguments (backend)

Required:

• --tokens: Path to the tokens.txt file for the speech model.
• --sense-voice: Path to the SenseVoice model.onnx.

Optional:

• --num-threads: Number of threads (default: 4).
• --microphone-index: Index of the microphone to use. If not specified, the system default microphone is used.
• --hotkey: Hotkey combination to start recording (default: ctrl+q).
• --api-port: Port number for the API server (default: 58652).
• --punc-model-dir: Directory path for punctuation model files (default: ./punc-onnx).
• --host: Host address for the API server (default: localhost).

---

HTTP API

The backend exposes a small HTTP API on --api-port (default 58652).

Base URL: http://localhost:58652

GET /ping

Health check to verify the backend is alive.

curl http://localhost:58652/ping

Response:

{"status": "alive"}

GET /list_audio_devices

List available audio input devices.

curl http://localhost:58652/list_audio_devices

Example response:

{
  "devices": [
    {
      "index": 0,
      "name": "Microphone (Realtek High Definition Audio)",
      "is_current": true
    },
    {
      "index": 1,
      "name": "Stereo Mix (Realtek High Definition Audio)",
      "is_current": false
    }
  ]
}

POST /set_audio_device

Set the current microphone by index.

curl -X POST http://localhost:58652/set_audio_device \
  -H "Content-Type: application/json" \
  -d '{"index": 1}'

POST /set_hotkey

Configure the hotkey used to start and stop recording.

curl -X POST http://localhost:58652/set_hotkey \
  -H "Content-Type: application/json" \
  -d '{"hotkey": "ctrl+q"}'

GET /get_hotkey

Retrieve the current hotkey configuration.

curl http://localhost:58652/get_hotkey

Example response:

{"hotkey": "ctrl+q"}

---

Building & Packaging

Package backend (Windows)

cd backend
./build_onefile.bat

Frontend dev & build

cd frontend
npm install

# Run in development
npm run build
npm start

# Package for distribution
npm run build
npm run electron:build

Create a full release folder

1. Copy everything from backend\dist.
2. Copy everything from frontend\release\win-unpacked.
3. Place them in the same directory.
4. Run SayKey.exe.

---

FAQ

Nothing happens when I hold Ctrl+Q.

• Check that SayKey is running (capsule icon is visible).
• Make sure your cursor is in a text field.
• Try switching to another microphone in the tray menu.

Windows says "protected your PC" and blocks the app.

• Click More info → Run anyway if you trust the binary from this repository.

---

Contributing

We welcome issues and pull requests.

1. Fork the repository.

2. Create a feature branch:

   git checkout -b feature/your-feature

3. Commit your changes and push the branch.

4. Open a Pull Request.

---

License

SayKey is MIT licensed. See LICENSE.txt for details.

---

Acknowledgements

• SenseVoice for the speech-to-text engine.
• CT-Transformer-punctuation by lovemefan for punctuation capabilities.