VocoType - Precise Offline Voice Input Method

VocoType is a completely free desktop voice input method designed for professionals who value privacy and efficiency. All recognition is performed locally, ensuring it works without an internet connection and never uploads any data.

This GitHub project is the CLI (Command Line Interface) open-source version of the VocoType core engine, primarily intended for developers.

➡️ Want the best experience? Download the free desktop version now!

Ready to use out of the box, with more complete features and no technical background required.

Visit the official website to download the free, full-featured VocoType desktop version

Introduction

VocoType is an intelligent voice input tool that allows you to convert speech to text in real-time and automatically input it into your current application using hotkeys. It supports MCP speech-to-text, AI text optimization, custom replacement dictionaries, and more, making voice input more efficient and accurate.

📹 Demo Video

Related MCP server: vibevoice-asr

Download

🤔 Why is VocoType different?

✅ Core Features

• Complete GUI: Ready to use out of the box, with clear and intuitive operations.

• System-wide Input: Voice input directly into any software or text box.

• Custom Dictionary: Support for adding 20 common terms and names to improve recognition accuracy.

• 100% Offline: Absolute privacy and data security.

• Flagship Recognition Engine: Accurate recognition of mixed Chinese and English content.

• AI Intelligent Optimization: Supports selecting multiple AI models to automatically correct typos, homophones, and self-corrections in transcriptions via customizable prompt templates. It intelligently recognizes correction commands in spoken language (e.g., "no, change to"), making the output text more accurate and fluent.

(For professional users with higher requirements, an option to upgrade to the Pro version is provided within the app to unlock advanced features like an unlimited dictionary.)

🎯 Suitable for Various Professional Scenarios

Whether you are a writer, lawyer, scholar, gamer, or just handling daily office work, VocoType is your trusted efficiency partner.

✨ VocoType Core Engine Features

All VocoType versions share the same powerful core engine.

• 🛡️ 100% Offline, Privacy Assured: All speech recognition is performed locally on your computer.

• ⚡️ Flagship Recognition Engine: Accurate mixed Chinese and English input, no more repeated editing.

• ⚙️ Highly Customizable: Unique replacement dictionary feature ensures names, locations, and industry terms are correct the first time.

• 💻 Lightweight Design: Requires only 700MB of RAM, pure CPU inference, no expensive graphics card needed.

• 🚀 0.1s Response Time: Experience the speed of "what you say is what you get," ensuring your inspiration is never interrupted by waiting.

🛠️ [For Developers] CLI Installation Guide

Please note: This version is intended for developers with some technical background. If you are not familiar with the command line, we strongly recommend visiting the official website to download the easy-to-use VocoType Free Desktop Version.

1. Environment Dependencies

• Python 3.12

• We strongly recommend using uv or venv to create a virtual environment.

2. Clone and Install

# 1. 克隆仓库
git clone https://github.com/233stone/vocotype-cli.git
cd vocotype-cli

# 2. (推荐) 创建并激活虚拟环境
pip install uv
uv venv --python 3.12
source .venv/bin/activate  # macOS/Linux
# 或者 .\.venv\Scripts\activate  (Windows)

# 3. 安装依赖
uv pip install -r requirements.txt

# 4. 运行
python main.py

# 保存数据集运行
python main.py --save-dataset

Model Download: Upon first run, the program will automatically download approximately 500MB of model files. Please ensure your network connection is stable.

🌐 Volcengine BigASR Streaming Recognition Backend (Optional)

In addition to the default local FunASR offline engine, VocoType CLI also supports integrating Volcengine Doubao Large Model Streaming Speech Recognition as a cloud recognition backend.

Advantages

Configuration Steps

1. Log in to the Volcengine Console, create a speech application, and obtain your App Key and Access Key.

2. Create a config.json in the project directory:

{
  "backend": "volcengine",
  "volcengine": {
    "app_key": "YOUR_APP_KEY",
    "access_key": "YOUR_ACCESS_KEY",
    "resource_id": "volc.bigasr.sauc.duration",
    "enable_punc": true,
    "enable_itn": true
  }
}

3. Start with the --config parameter:

python main.py --config config.json

Note: When using the Volcengine backend, recording data will be sent to Volcengine servers for recognition and will no longer be completely offline. If you have strict privacy requirements, please continue using the default local FunASR backend.

FAQ

Q: Is my data safe?

A: 100% safe. All speech recognition is performed locally offline; your audio data will not be uploaded to any server.

📞 Contact Us

• Bugs and Suggestions: Please use GitHub Issues as a priority.

• Follow us for the latest updates: https://vocotype.com

🙏 Acknowledgments

The birth of VocoType would not have been possible without the following excellent open-source projects:

• FunASR - An open-source speech recognition framework from Alibaba DAMO Academy, providing VocoType with powerful offline speech recognition capabilities.

• QuQu - An excellent open-source project that provided important technical references and inspiration for VocoType.

Thanks to these open-source communities for their selfless contributions!