AI Deck

Handy — A fully offline, open-source voice input app that turns speech into text with a single shortcut

A free, open-source desktop dictation app developed by CJ Pais. Hold a keyboard shortcut, speak, and a speech recognition model transcribes your words and inserts the text straight into whatever field you have open. Everything runs on your own machine ─ no audio is ever sent to the cloud. You can choose between two model families: Whisper, which is strong across many languages, and Parakeet, which is fast for English and European languages. Built with Tauri, it runs on macOS, Windows, and Linux under the MIT license, and can be installed via Homebrew, winget, or a Raycast extension.

Key Features

  • Works fully offline: Speech recognition models are downloaded to your device, so neither the recorded audio nor the transcript leaves your machine. It works on a plane, in an offline environment, or in workflows where audio must not leave the organization
  • Press and speak: The default shortcut is Option + Space. You can use push-to-talk, which records only while the key is held, or a toggle mode that starts and stops recording with each press. Escape cancels
  • Works in any text field: The transcript is inserted directly into the active text field ─ mail clients, chat apps, code editors, browser forms, and anything else
  • Whisper and Parakeet for different needs: Use Whisper when you need coverage of 99+ languages, and Parakeet when you want fast, accurate English or European language transcription. Whisper is available in Small, Medium, Turbo, and Large variants
  • GPU acceleration and silence removal: Whisper models are accelerated where a GPU is available (NVIDIA, AMD, and Intel GPUs on Windows and Linux). Silero VAD strips out silent stretches, which keeps pauses and hesitations from bloating the audio
  • AI post-processing: Option + Shift + Space runs “transcribe with post-processing,” letting an AI clean up and reformat the text. A custom words feature lets you register proper nouns you use often
  • Multiple installation paths: Download directly from the official site or the GitHub releases page, or install via brew install --cask handy on macOS and winget install cjpais.Handy on Windows. Raycast can launch it as well

Pricing

PlanPriceHighlights
Open source$0Full functionality with no limits. MIT licensed, no account or subscription required

The developer states that accessibility tooling belongs in everyone’s hands rather than behind a paywall, so there are no paid tiers or feature gates. The only cost is your device’s own compute ─ there are no API fees.

Pricing information is as of August 2026. Please check the official website for the latest pricing.

Pros & Cons

Pros

  • Audio never leaves your machine, so it is safe to use for dictation containing confidential or personal information
  • Free with no subscription. Costs do not grow no matter how much you talk
  • Works the same way on an unstable network, or with no network at all
  • MIT-licensed open source, so you can inspect the behavior and extend it yourself
  • Supports macOS, Windows, and Linux, so it is not tied to one platform

⚠️ Cons

  • Since models run locally, weaker machines will see longer waits. Larger models require downloads exceeding 1 GB
  • Issues with Bluetooth headsets on macOS have been reported
  • On Linux under Wayland, text insertion requires additional tools such as wtype or xdotool
  • There are known crashes with Whisper models on some Windows and Linux configurations
  • Initial setup requires granting microphone and accessibility permissions and downloading a model
  • As an individually maintained project, there is no enterprise support desk or SLA

Comparison with Similar Services

CriteriaHandyMacWhispersuperwhisperWispr Flow
Supported OSmacOS / Windows / LinuxmacOSmacOSmacOS / Windows
Where processing happensOn device (fully offline)Primarily on devicePrimarily on deviceCloud
Pricing modelFree (open source)Free tier + paid licenseFree tier + paid plansSubscription
Source codePublic (MIT)ClosedClosedClosed
Main useDictation into any appDictation + audio file transcriptionDictation + AI formattingDictation + AI formatting

Pricing and features change for all of these services, so check each official site for current details when comparing.

Who Is It For

  • People who dictate sensitive content and do not want audio sent to the cloud
  • Anyone who would rather speak long emails or documents than type them
  • Daily dictation users who want to avoid monthly or usage-based billing
  • Windows or Linux users for whom macOS-only apps are not an option
  • Anyone who needs voice input for accessibility reasons ─ reducing strain on hands and shoulders, or difficulty with keyboard input
  • Developers who want to inspect and modify the tools they rely on

Summary

Handy narrows dictation down to a single idea ─ press and speak ─ and delivers it fully offline, free, and open source. Compared with cloud dictation services, it asks for some upfront effort in downloading models and granting permissions, and speed depends on your hardware. In exchange, you get the assurance that your audio stays put and the freedom of costs that never grow with use. Start with a lightweight Parakeet or Whisper Small model, and move up to a larger one if accuracy falls short.

← Blog