Rank #10 · Daily · Multimedia
Handy: Free Offline Speech-to-Text Desktop App (Whisper / Parakeet)
Press a shortcut, speak, get text pasted into any field—entirely offline, extensible, and free.
What it does
Handy is a cross-platform desktop speech-to-text application built with Tauri/Rust. Default loop:
- Press a configurable shortcut (hold-to-record, toggle, or hold-/toggle-only modes).
- Speak while recording is active.
- On stop, Silero VAD filters silence; Whisper (Small/Medium/Turbo/Large, GPU when available) or Parakeet (CPU-optimized, auto language detection) transcribes locally.
- Text pastes into the focused app.
Extras for power users: optional LLM post-processing; macOS/Linux hooks/transcription executable for last-mile rewrite before paste; CLI flags (--toggle-transcription, --start-hidden, etc.) useful especially on Wayland; Raycast extension; manual model install paths for proxy/restricted networks; custom Whisper GGML discovery; release artifact verification via Tauri updater signatures / minisign.
Install from GitHub Releases or handy.computer; also Homebrew cask and winget (community-maintained packages noted). Grant microphone and accessibility permissions on first run. Linux needs xdotool / wtype / dotool (and often libgtk-layer-shell) for reliable paste—README documents compositor caveats.
Who it’s for
Anyone who wants privacy-first dictation without a SaaS subscription—writers, developers, accessibility users—on Windows, macOS, or Linux. Compared with Hex (#07), Handy is the more documented, feature-rich offline STT desktop with model choice and extension surface.
Stars / language / license
~37,041 stars · Rust · MIT
Data verified: 2026-11-02 ~22:16 Asia/Shanghai (GitHub API get_repository + README get_file_contents)
Trending note (selection ~22:14 Asia/Shanghai): ~+339 stars that day (archive); live API confirms a still-active high-star desktop STT repo
Category tags
Multimedia · speech-to-text · offline · Whisper · Parakeet · Tauri · accessibility · privacy
How it differs from similar projects
Versus cloud dictation lock-in, Handy’s default is offline hotkey-to-paste. Versus Hex on the same Daily list, Handy exposes a larger product surface (models, history, dictionary, hooks, Raycast, signed releases, troubleshooting depth). Versus raw Whisper/ggml CLIs, it is a shipped desktop UX with VAD and paste automation across three OSes. Code is MIT; the Handy name, logo, and brand assets are not open-source—forks must rebrand. No third-party accuracy bake-off is claimed here; try models on your accent and hardware.
Practical caveats (from the README)
- Bluetooth headsets on macOS may degrade playback while recording (bidirectional Bluetooth); prefer built-in/external mic with headphones as output.
fn/ Globe shortcuts only work on Apple keyboards—use standard modifiers on third-party boards.- Linux overlay can steal focus; default overlay off or set Overlay Position to None; enable audio feedback if you still want recording cues.
- Clipboard paste races under load: History may show correct text while an older clipboard pastes—debug menu paste delays / Reliable Paste (beta) are documented workarounds.
None of these are deal-breakers for most users; they are the difference between a demo and daily-driver STT.