What is Handy?
Handy, by CJ Pais, is a desktop app built in Rust on Tauri that turns speech into typed text. You press a shortcut (Option and Space on a Mac by default), speak, and press it again or let go; Handy trims silence with the Silero voice-activity model, transcribes the clip on your own computer and pastes the result into the focused app. It needs microphone access, and on macOS accessibility permission so it can paste into other apps.
You pick the speech model on first launch. Parakeet V3 is the recommended default for English and European languages and runs well on a CPU; Whisper Small, Medium, Turbo and Large cover 99 or more languages and use the GPU when one is present; Cohere Transcribe is offered for the best English quality on a capable GPU. Optional extras include a transcript history with audio playback, a transcription hook that rewrites text before pasting, command-line flags and an alpha post-processing step that passes text through a language model.
The code is MIT-licensed, though the Handy name and logo are not. The repository dates from February 2025 and has about 39,389 GitHub stars and 3,719 forks; v0.10.1 shipped on 14 November 2026.
Is Handy worth it?
Handy's strength is that it stays small. It does not take meeting notes, rewrite your tone or keep a cloud account; it records a short clip while you hold a key, transcribes it locally and pastes it. The FAQ says the only network requests are model downloads and update checks, which makes it one of the few dictation tools a privacy-minded team can approve without a vendor review.
The model choice is where you trade speed for range. Parakeet V3 is fast on ordinary laptops and is the default for English and European languages. Whisper reaches 99 or more languages but is heavier: the downloads run from 487 MB for Small to 1,600 MB for Turbo. Cohere Transcribe is the pick for English accuracy if you have a GPU. Against Wispr Flow, the paid leader, Handy gives up automatic clean-up of filler words and per-app formatting unless you switch on post-processing.
That post-processing is the one place your words can leave the machine. It is an alpha feature behind an experimental toggle with its own shortcut; with OpenAI, Anthropic, OpenRouter, Groq, Cerebras or Z.AI selected, the transcript goes to that provider under your API key. Apple Intelligence on macOS 26 or a local OpenAI-compatible endpoint keeps it on the device. History is local too, but it stores each recording as audio on disk, so set a limit or 0 if you dictate anything sensitive.
The catch: pasting into other apps is the fragile step. Windows will not let it paste into apps running as administrator, Wayland support on Linux is limited and needs wtype, dotool or ydotool, and under load some apps read the clipboard after Handy has restored it, so you get the text you copied earlier. Handy transcribes only the microphone, not system audio, and does not stream live captions; it waits until you stop speaking.
Our take: for dictation into a browser, editor or chat window, Handy is the free option to install before paying for anything, and Parakeet V3 is the model to start on. People who want their speech tidied as they talk will still find Wispr Flow's polish worth the fee.
Where it earns its place
A developer with repetitive strain injury
Write commit messages, chat replies and prompts for a coding agent without typing
Install Handy, pick Parakeet V3, and hold the shortcut while speaking into the terminal or editor. On Linux, use the Shift+Insert paste method so it works in terminals too. No cost and no audio leaves the laptop.
A solicitor's office with client-confidentiality rules
Dictate file notes without sending audio to a cloud service
Deploy Handy with post-processing left off and History set to 0 or a short limit. Transcription stays on each PC; the remaining risk is the operating system's clipboard, which Handy restores after pasting.
A support agent who answers tickets in English and Spanish
Draft replies by voice in either language
Choose Whisper Medium (492 MB) for multilingual input and turn on post-processing with a local model through LM Studio to fix punctuation, so the text still stays on the machine.
Before you commit
- Setup
- Download the installer from handy.computer or GitHub (Homebrew and winget packages are community-maintained), grant microphone and accessibility permissions, and download a model on first launch. Minutes to the first transcript.
- Integrations
- Pastes into any focused text field; CLI flags and Unix signals for window-manager shortcuts; a Raycast extension; a transcription hook on macOS and Linux; post-processing through OpenAI, Anthropic, OpenRouter, Groq, Cerebras, Z.AI, Apple Intelligence or any OpenAI-compatible endpoint.
- Data and privacy
- Transcription is local. Network use is model downloads and update checks, unless you enable cloud post-processing, which sends the text to that provider. History stores audio recordings in the app data folder; a limit of 0 disables it.
- Cost at scale
- Free per seat. The only running cost is API usage if a team turns on cloud post-processing.
- Limits
- Records a segment, then transcribes, with no streaming. Hook runs time out after 30 seconds. fn-key shortcuts work only on Apple keyboards. No paste into Windows administrator apps.
What’s good
- Transcribes on your computer; only model downloads and update checks go online
- Windows, macOS and Linux, free under MIT
- Choice of Parakeet V3 on CPU, Whisper on GPU or Cohere Transcribe
- Signed releases, CLI flags and a hook to rewrite text before pasting
What’s not
- Cannot paste into Windows apps running as administrator
- Limited Wayland support; Linux needs an extra typing tool
- No system audio and no live captions
- History keeps audio recordings on disk unless you limit it
Best for
Anyone who dictates into ordinary text fields and wants it free and offline, especially on Linux or Windows where good options are scarce.
Handy alternatives
Handy: common questions
- Is Handy worth it?
- Handy does the one thing most paid dictation apps charge a subscription for, offline and free, and does it on all three desktop systems. We rate it 4.3 out of 5.
- How much does Handy cost?
- Free and open source under MIT, with optional donations and GitHub Sponsors. Optional cloud post-processing uses your own API key with that provider's prices.
- What are the downsides of Handy?
- Cannot paste into Windows apps running as administrator; limited Wayland support; Linux needs an extra typing tool; no system audio and no live captions; history keeps audio recordings on disk unless you limit it.
- Who is Handy best for?
- Anyone who dictates into ordinary text fields and wants it free and offline, especially on Linux or Windows where good options are scarce.
- How much does Handy cost for a team or in production?
- Free per seat. The only running cost is API usage if a team turns on cloud post-processing.
- Is Handy safe for company data?
- Transcription is local. Network use is model downloads and update checks, unless you enable cloud post-processing, which sends the text to that provider. History stores audio recordings in the app data folder; a limit of 0 disables it.
- What are the best Handy alternatives?
- The closest alternatives we have reviewed are VoiceStudio and Whisper (openai/whisper).
Answers checked Nov 20, 2026.