Back to catalog
Whis
MITWhis is an open-source voice-to-clipboard dictation tool for Linux that replaces proprietary speech services with local privacy and custom AI presets.
Overview
A desktop user drafting emails or asking an AI assistant questions often spends more time typing and correcting mistakes than thinking. Whis translates spoken voice into text and drops it straight into the system clipboard upon pressing a hotkey. It bridges speech recognition and desktop apps without expensive software subscriptions or vendor lock-in.
Where Whis fits
Commercial voice dictation utilities like Wispr Flow or Superwhisper provide speech entry on macOS and Windows, but they depend on closed background daemons, mandatory user accounts, and monthly fees. Linux users are frequently left out or forced to stitch together custom scripts. Whis fills this gap with a dedicated speech-to-clipboard pipeline designed for Linux, supporting cloud transcription endpoints alongside offline local processing.
Operationally, the differences from proprietary services are stark. Whis runs without background telemetry daemons or cloud tie-ins, giving users direct control over their stack. For cloud transcription, it connects directly to provider APIs like Groq, OpenAI, or Deepgram using personal developer keys, avoiding per-seat markups. For environments requiring strict privacy, Whis runs local models on device via an embedded speech engine. However, this flexibility shifts setup to the user: you must configure desktop shortcuts, supply API keys or model weights, and maintain audio libraries like ffmpeg.
What it does well
- Direct clipboard capture: speaking into a microphone and toggling a shortcut places transcribed text onto the clipboard, ready to paste into any browser or editor.
- Pluggable processing backends: users can choose between fast cloud speech services using personal API keys or private local models that run entirely offline.
- Desktop tray integration: an optional graphical interface provides a system tray icon with clear recording indicators, avoiding terminal windows.
- Intelligent prompt presets: built-in transformation rules remove verbal pauses and format spoken thoughts into structured emails, notes, or AI prompts.
- Prerecorded audio transcription: the utility transcribes audio files directly from disk to produce written drafts of lectures or voice memos.
- Minimal background footprint: written in Rust, the daemon consumes negligible memory and processor cycles while waiting for shortcut triggers.
