QuQu - An open-source desktop voice input and text processing tool

What is a cricket (QuQu)?

QuQu is a desktop voice input and text processing tool designed for Chinese users, and is an open-source and free alternative to Wispr Flow. The tool incorporates the Alibaba FunASR Paraformer model, supports local operation, protects user privacy, and combines AI technology to achieve accurate recognition, intelligent optimization, and context-aware text output. QuQu supports programming terminology recognition, is compatible with various top domestic AI models, and features one-click wake-up, real-time recognition, and seamless pasting, allowing users to efficiently use voice input in quiet environments, freeing their hands and improving work efficiency.

The main functions of crickets (QuQu)

• One-click wake-upYou can start voice input anytime, anywhere using global shortcut keys, which is convenient and fast.
• Real-time recognitionThe local FunASR engine provides high-precision Chinese speech recognition, maintaining high accuracy even in quiet environments.
• Intelligent optimizationIt connects to AI models to automatically polish text, correct slips of the tongue and erroneous expressions, and improve text quality.
• Seamless bondingThe converted text is automatically pasted to the current cursor position without manual operation.
• Context awarenessSupports context-aware intelligent output, automatically adjusting text content based on the context.
• Programming supportIt supports programming terminology recognition, making it suitable for developers and efficiency experts.
• Privacy protectionIt supports local operation and does not upload data to the cloud, ensuring user privacy and security.
• Multi-platform supportCompatible with macOS, Windows, and Linux systems.

The technical principles of crickets (QuQu)

• Speech recognition engine:
  • FunASR ParaformerIt uses Alibaba's FunASR model, especially the Paraformer-large model, to provide high-precision Chinese speech recognition. The model can run locally, ensuring data privacy.
  • FSMN-VADUsed in speech activity detection, it can accurately identify the beginning and end of speech segments.
  • CT-TransformerUsed in text decoding to improve the accuracy and efficiency of recognition.
• AI model integration:
  • Two-stage engineIt combines ASR (Automatic Speech Recognition) and LLM (Large Language Model) technologies to achieve accurate recognition and intelligent optimization of voice input.
  • Compatible with OpenAI APIIt supports a variety of services compatible with the OpenAI API, and prioritizes adaptation to top domestic models, such as Alibaba Cloud Tongyi Qianwen and Kimi.
• Front-end and desktop technology stack:
  • front end: Use React 19, TypeScript, Tailwind CSS, shadcn/ui, and Vite to build the user interface and provide a smooth user experience.
  • DesktopBuilt on Electron, ensuring cross-platform compatibility.

QuQu's project address

• GitHub repositoryhttps://github.com/yan5xu/ququ

Applications of QuQu (a type of cricket)

• Writing and Content CreationIt quickly converts speech to text, improving writing efficiency and is suitable for writers, bloggers, journalists, and other scenarios that require a large amount of text input.
• Programming aidsIt supports programming terminology recognition, helping developers quickly input code, reduce keyboard input, and improve programming efficiency.
• Meeting minutesReal-time speech-to-text conversion, automatic editing and summarization, facilitating the organization of meeting minutes and keynotes.
• Study and NotesQuickly jot down key points in class or lectures, reducing the burden of handwritten notes.
• Office document processingIn a quiet office environment, you can complete document writing and editing via voice input, improving work efficiency.