High-Performance Offline Chinese Speech Input Method for Linux.
🔗 View on GitHub
VocoType-linux is an open-source, high-performance offline Chinese speech input method designed for modern Linux desktops. It brings intuitive and efficient voice typing to Linux by combining a powerful speech recognition engine with native input method integrations.
Unlike typical cloud-based voice input solutions, VocoType-linux runs entirely locally, ensuring privacy and data safety since no audio or text data is uploaded to external servers. Its core speech recognition is powered by the FunASR Paraformer model, providing fast (~0.1 s), accurate Chinese (and mixed-language) transcription with minimal resource usage.
The project supports the two major Linux input frameworks, IBus and Fcitx5, so users on GNOME, KDE, and other distributions can install and use it seamlessly within their existing desktop workflows. Both versions share the same recognition engine but are adapted to their respective input framework protocols for stable and responsive performance.
Key highlights of VocoType-linux:
- 100% offline speech recognition — no internet required, full privacy protection.
- Fast response — voice input commits text in around 0.1 seconds after speech.
- Low resource usage — optimized for CPU-only inference and modest memory footprint (~700 MB).
- Broad Linux support — works with both IBus and Fcitx5 input method frameworks.
- Flexible usage scenarios — suitable for chatting, document writing, development comments, emails, and more.
VocoType-linux is ideal for developers and privacy-conscious users who want voice input on Linux without relying on cloud services.