Vocalinux

Vocalinux is a voice dictation tool built specifically for Linux desktops. Hold Right Alt, speak, and the transcript appears in whatever app you're already using: a terminal, browser, IDE, or office document. No cloud account, no usage telemetry, no mandatory internet connection.

It lives in the system tray and stays out of the way until you need it. Toggle mode is available if you'd rather not hold the key. IBus handles text injection when available, with a clipboard fallback for Wayland compositors that don't bridge input.

Engines you can choose from:

• whisper.cpp (default): C++ Whisper with Vulkan GPU support across AMD, Intel, and NVIDIA hardware. The tiny model is around 74 MB.
• Faster Whisper: a faster Python-based Whisper variant for those who want it.
• VOSK: CPU-friendly streaming with models around 40 MB, suited for older or low-RAM machines.
• Whisper: the original OpenAI PyTorch path, best for NVIDIA CUDA setups.
• Parakeet: another on-device option in the engine list.
• Remote API: point it at an OpenAI-compatible server you control. Local VAD still applies; audio goes to your server, not a Voca cloud.

All local engines process audio entirely on your machine. The app doesn't send usage pings or analytics. If you configure a Remote API endpoint, audio travels only to the host you specify.

Unlike macOS-focused tools such as VoiceInk or Superwhisper, Vocalinux is built around Linux desktop realities: X11 and Wayland support, IBus integration, and packages for Ubuntu, Fedora, Debian, Arch, and openSUSE. Native .deb, .rpm, Snap, and Flatpak packages are available. The interactive installer detects your hardware and lets you pick an engine before wiring up the desktop app.

Licensed under AGPL-3.0 with no premium tiers.