Home

TypeWhisper for Linux

Speech-to-text and AI text processing for the Linux desktop. Press a key, talk, and have clean, punctuated text land in whatever app you're in.

TypeWhisper for Linux is a Linux desktop port forked from the TypeWhisper project (which ships macOS and Windows builds). It lets you dictate into other applications, transcribe audio files, record longer WAV sessions, apply dictionary and snippet post-processing, and run prompt-based AI text actions through plugins — tuned to feel as close to Wispr Flow as possible on Linux.

New here? Start with these

• Installation — AppImage, .deb, .rpm, tarball, or build from source.
• Quick Start — your first dictation in a few minutes.
• Setup Wizard — the guided first-run setup; reopenable anytime.
• Requirements — system requirements and optional desktop helpers.
• On Wayland? Read Wayland Notes — hotkeys, typing backends, window/URL detection.

What it does

• Global dictation with Toggle, Push-to-talk, and Hybrid activation modes — see Dictation and Global Hotkeys.
• Local transcription on CPU or optional NVIDIA GPU — CUDA acceleration (for the bundled whisper.cpp and sherpa-onnx engines) downloads its runtime on demand rather than bundling it. See Transcription Engines and GPU & CUDA.
• Per-app text insertion with smart Auto, clipboard paste, and direct typing — see Text Insertion.
• AI text cleanup & formatting in the Wispr-Flow style, driven by your own LLM — see Text Cleanup & Formatting and Prompts & AI Actions.
• Voice-native editing — start a dictation with a keyphrase and the rest becomes a command that edits your highlighted text or writes new text at the cursor. Ships disabled; see Prompts & AI Actions.
• File transcription and a recorder, with batch queues, watch folders, and subtitle export — see File Transcription and Recorder.
• Personalization via History, Dictionary, Snippets, and Profiles.
• Learns from your corrections — type over a dictated word in the target app to fix it and TypeWhisper silently learns the correction, then applies it to future dictations. Off by default; see Dictation.
• Plugins for transcription engines, LLM providers, actions, post-processing, memory, and notifications — see Plugins.
• Automation through a local HTTP API and the CLI.
• A localized interface — English, German, Spanish, or Russian, switched live from General Settings (or Auto, to follow your system locale).

Map of the wiki

Section	Pages
Getting Started	Installation · Quick Start · Setup Wizard · Requirements
Using TypeWhisper	Dashboard · Dictation · Global Hotkeys · Text Insertion · File Transcription · Recorder · History · Dictionary · Snippets · Profiles · Prompts · Text Cleanup · Memory
Settings	General Settings · Appearance · Advanced Settings · About · Backup & Restore
Plugins	Plugins · Transcription Engines · LLM Providers · Other Plugins · Plugin SDK
Automation	HTTP API · CLI
Platform & Troubleshooting	Wayland Notes · GPU & CUDA · Desktop Integration · Tested Configurations · Troubleshooting
Project	Data & File Paths · Roadmap · Contributing · License

---

Changelog

Date	Change
2026-06-17	Initial version.
2026-06-17	Added Dashboard hero screenshot.
2026-06-17	Linked the Setup Wizard page.
2026-06-19	Added the localized-interface highlight (v0.9.0).
2026-06-23	Added the local GPU-acceleration highlight (v0.10.0: on-demand CUDA for whisper.cpp and sherpa-onnx).
2026-06-24	v0.11.0: hardened the on-demand GPU runtime — corrupt-cache recovery (Clear GPU runtime), resumable downloads, and smarter CPU/GPU restart prompts.
2026-07-08	v0.12.0: added spoken command mode (voice-native editing via a keyphrase) and the History Inspect panel (opt-in LLM prompt/response transparency).
2026-07-27	v0.12.1: documented spoken-number normalization, the Automatic (follow system default) microphone, and data-preserving uninstall (--purge to opt out).
2026-07-27	v0.13.0: added the correction-learning highlight, plus session-lock, paste-ordering, restore-validation, and atomic-write behavior across the platform pages.