Hex
Hex is a free, MIT-licensed voice dictation app for macOS, Windows and Linux by Christopher Langton. You hold a key or tap it, speak, and the text is pasted into the app at your cursor. Its README says the first run downloads a speech model and after that nothing leaves your machine. It bundles Parakeet, Whisper, Qwen3-ASR, SenseVoice, Nemotron and Granite as choices.
What does Hex do?
Hex's README calls it "local voice dictation for Mac, Windows and Linux. Hold a key, speak, and it pastes." You hold the shortcut to dictate, or set Shortcut mode to Tap to start and stop with a press, and "the text lands at the cursor in whatever app you're in." The README says "Hex transcribes on your device and pastes into whatever app you're in" and that "the first run downloads a speech model; after that nothing leaves your machine."
The app also keeps a seven-day searchable history, runs each dictation through a chain of output steps you choose (word replacements, built-in cleanups, your own scripts, an optional OpenCode rewrite), biases transcription toward a vocabulary list you set, and ships a local API with a TypeScript SDK and a command-line tool.
Key facts
- Licence: MIT, verified from the LICENSE file (2026 Christopher Langton).
- Built-in speech models: Parakeet Unified English, Parakeet v3 (25 European languages), Cohere Transcribe (14 languages), Whisper large-v3-turbo, Qwen3-ASR 0.6B (Chinese and 29+ languages), SenseVoice Small, Nemotron 3.5 ASR (27 languages), Granite Speech 5.0 (7 languages), plus single-language models including ivrit.ai Whisper, GigaAM v3, NB-Whisper and QwenCleo-ASR.
- Custom GGUF models are also supported.
- Install: Homebrew cask (
brew install --cask anomalyco/tap/hex), AUR (yay -S hex-bin), a Linux install script, Nix, a DMG for macOS, installers for Windows and.deband.rpmpackages for Linux. - Dictation history is kept for seven days by default.
- Vocabulary biasing: Whisper and Qwen3-ASR read the list while transcribing; Parakeet reads it with Labs then Vocabulary biasing on.
- Annotate records the screen while you draw and talk, on macOS, Linux and Windows (GNOME Wayland needs Labs then Annotate on GNOME).
- Meetings is a macOS Labs feature that records a call and writes the transcript.
- The CLI (
hex transcribe meeting.m4a) prints a file's transcript whether or not the app is running, and reads WAV, AIFF, CAF, FLAC, MP3, Ogg, WebM and MP4/M4A/MOV.
What does it replace, and where does it fall short?
Hex replaces Wispr Flow and Superwhisper for someone who wants a dictation key on their laptop, the recognition on their own machine, no subscription and no account. It is listed on the open-source alternatives to Wispr Flow page.
Where it falls short: the README does not promise an iPhone or Android build, so a phone user has no path. Dictation history is seven days by default. The Meetings feature is macOS only, and Annotate on GNOME Wayland needs a Labs toggle. The README documents no minimum hardware, so a reader cannot tell before installing whether a given laptop will keep up with a given model.
How does Hex run?
As a desktop app you install from the project's own packages or from Homebrew, the AUR, a Linux script or Nix. The first run downloads a speech model (you pick which one), and after that the recognition runs on the machine. You hold a shortcut to dictate, or set Tap mode and press once to start and once to stop. Option-Shift-V (Ctrl-Alt-V on Windows and Linux) pastes the last dictation again.
For teams that want to script Hex, an authenticated local API and a TypeScript SDK transcribe with the same model, and a command-line tool (hex transcribe, hex status, hex dictation, hex settings) drives the running app and transcribes files.
Who is Hex for?
Someone who wants a dictation key that works offline after the first model download and who wants to pick the model themselves, from English-only Parakeet to multilingual Whisper to Qwen3-ASR for Chinese. A reader who wants a cleaner polished-prose result should also look at VoiceInk, which adds a language-model step; a reader who wants the smallest install should look at Handy.
What limits does the README state?
The README does not have a dedicated "known issues" section. It does say dictation history is seven days by default, that Meetings is a macOS Labs feature, and that Annotate on GNOME Wayland needs the Labs toggle on. The README does not state minimum hardware for any model.
Questions people ask
Is Hex open source?
Yes. Hex is MIT, verified from the LICENSE file: "Copyright (c) 2026 Christopher Langton. Permission is hereby granted, free of charge, to any person obtaining a copy of this software..." MIT is an OSI-approved licence with no added conditions, so you can read, change and ship your own build.
Does Hex send my voice to the cloud?
The README says "the first run downloads a speech model; after that nothing leaves your machine." Transcription runs on the device. Cohere Transcribe is listed as one of the model choices and is a cloud model, so if you pick that one your audio does leave the machine; the Parakeet, Whisper, Qwen3-ASR, SenseVoice, Nemotron and Granite choices run locally.
Can Hex run offline?
Yes, once a local model is downloaded. The first run fetches whichever model you pick, then every dictation runs on the device. The CLI works whether or not the app is running, so a batch job on a file does not need a network either, except if you have chosen a cloud model as your default.
How does Hex compare with Wispr Flow?
Hex does the core of Wispr Flow (hold a key, speak, paste) with no subscription, no account and the recognition on your machine. Wispr Flow's product includes a polished rewrite step; Hex's output chain lets you add your own rewrite (word replacements, built-in cleanups, your own scripts, or an OpenCode rewrite). The open-source alternatives to Wispr Flow page sets it beside Handy, VoiceInk and OpenWhispr.
What does Hex offer that a simpler tool does not?
Vocabulary biasing on a named list, per-app output steps, screen annotation on all three desktops, meeting transcription on macOS, a seven-day searchable history, and a local HTTP API with a TypeScript SDK and a CLI. A reader who only wants hold-a-key-and-paste should pick a smaller tool; a reader who wants to script dictation into other software should look here first.
Sources
- Hex README and LICENSE: github.com/anomalyco/hex, read 2026-11-16. Code licensed MIT (Copyright 2026 Christopher Langton).
More voice dictation apps
A free, offline push-to-talk dictation app for macOS, Windows and Linux that runs Whisper or NVIDIA Parakeet on your own machine and pastes the text at your cursor, under MIT.
A GPL-3.0 macOS dictation app that runs Nemotron, Parakeet, Cohere Transcribe, Apple Speech or Whisper on the device, with command and write modes, per-app prompts, and an optional closed local enhancement model.
A dictation and meeting-notes app under MIT that runs Whisper or Parakeet locally or cloud models with your own key, detects Zoom, Teams and FaceTime calls, and labels speakers on device.
A macOS dictation app under GPL-3.0 whose README names Superwhisper and Wispr Flow as the tools it replaces, with local models, per-app modes, a personal dictionary and an AI assistant mode.
A speech-to-text app inside the Epicenter monorepo under AGPL-3.0 that records, transcribes with a provider you choose, cloud or local, optionally polishes the text, and pastes it at the cursor.
An Apache-2.0 macOS dictation app that calls Gemini 3.5 Transcribe with your own key, writes audio to disk from the first millisecond for crash safety, and inserts polished text at your cursor without an account.
Added November 16, 2026. Every claim above comes from the project's README, LICENSE or model card, read on November 16, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/anomalyco/hex.
Want Hex running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.