yap
yap is a free, open-source dictation app for macOS, Windows and Linux under MIT. Hold a key, speak, and what you said lands at the cursor in any app. It runs faster-whisper locally by default with no cloud call and no account. You can point it at an OpenAI-compatible cloud endpoint or a self-hosted Whisper server, and turn on an optional LLM cleanup pass.
What does yap do?
yap describes itself in its README as "free voice dictation, everywhere". You hold a key, speak, and the text is pasted at the cursor in your editor, browser, chat box, terminal or anywhere else you were typing. Its README says transcription runs locally with Whisper by default, so there is no account, no cloud, no telemetry and nothing screenshots the screen. Audio only leaves the machine if you deliberately switch to a cloud engine.
yap learns as you use it. yap vocab add biases Whisper toward a word; yap vocab fix is a guaranteed find-and-replace; and the "relearn hotkey" reads your last correction from the clipboard and saves the differences (casing goes to the glossary, spelling becomes a replacement). Per-app profiles merge extra settings over the base config while you dictate into a named app.
Key facts
- Licence: MIT, per the LICENSE file.
- Platforms: macOS, Windows and Linux, all from one config file. Requires Python 3.9 to 3.13 for the pipx install path.
- Default hotkey: hold Right Option (⌥) on macOS, Right Ctrl on Windows and Linux, push-to-talk. Toggle mode is optional.
- Engines:
localruns faster-whisper on the CPU or GPU with no cloud call;cloudsends audio to any OpenAI-compatible/audio/transcriptionsendpoint using your key. Groq and self-hosted Whisper are both examples in the README. - Local models:
auto(adapts to the machine),tiny.en,base.en,small,medium,large-v3. - Optional LLM cleanup step for punctuation and filler removal, against any OpenAI-compatible API using your key.
- Install:
pipx install "yap-dictation[full] @ git+https://github.com/AkuchiS/yap", or clone plus./install.shon macOS and Linux. A PyInstaller build ships as a signed macOS.appand a Windows.exe. - Auto-learning of repeated proper nouns is on by default, capped, and configurable under
learning. - Wayland Linux support runs through the compositor binding to a local yap socket (Hyprland, GNOME, KDE), because Wayland blocks apps from grabbing global hotkeys.
What does it replace, and where does it fall short?
yap replaces Wispr Flow and Superwhisper for someone who wants push-to-talk dictation that runs locally, on all three desktops, with no account and no meter. The README's comparison table says yap has no word limit and works offline by default. It is a candidate on the open-source alternatives to Wispr Flow page and the open-source alternatives to Superwhisper page.
Where it falls short: Wispr Flow ships an iOS and an Android app, and yap does not; the README names Mac, Windows and Linux only. The README's table says Wispr Flow screenshots the active window "for context" and yap never does, so anyone who wants that context step will not get it here. The macOS .fn and Globe key are not usable as hotkeys, per the README's note. For an alternative that also does LLM cleanup on macOS, see Whispree or VoiceInk.
How does yap run?
pipx install "yap-dictation[full] @ git+https://github.com/AkuchiS/yap" && yap run is the recommended install, per the README. The macOS build is signed and notarised. Linux users need xclip or xsel on X11 for clipboard fallback and wl-clipboard on Wayland. Windows needs no extras.
The daemon runs in the background with yap run, and yap app adds a menu-bar or system tray icon. yap doctor prints diagnostics for permissions, the hotkey, the microphone and the clipboard. There is no hosted paid version and no account.
Who is yap for?
A person who wants Wispr-style dictation on all three desktops without a subscription and without sending audio anywhere by default. A person who wants iOS or Android dictation should stay on Wispr Flow.
What limits does the README state?
Python 3.14 is too new for some native dependencies. The macOS .fn and Globe keys emit no real keypress and cannot be used as the hotkey (Right Option is the workaround). Wayland deliberately blocks apps from grabbing global hotkeys, so the fix is to bind the key in the compositor (Hyprland, GNOME or KDE) and let it poke yap toggle on a local socket. Some Wayland compositors also restrict synthetic keystrokes, so paste may need inject.method = "type".
Questions people ask
Is yap open source?
Yes. The LICENSE file is MIT, an OSI approved licence, with no additional restriction. The README repeats that yap is "MIT-licensed, with no meter and nothing to upsell". Anyone can read, modify and redistribute it, including for commercial use.
Does yap send my voice to the cloud?
Not by default. The README says the default engine is local, running faster-whisper on your CPU or GPU, and audio never leaves the machine. If you switch engine to cloud and set an YAP_API_KEY, audio is sent to the OpenAI-compatible endpoint you configured (Groq, OpenAI, or a self-hosted server); that is a deliberate opt-in.
Which platforms and hotkeys does yap support?
macOS, Windows and Linux, per the README. The default hotkey is Right Option on macOS and Right Ctrl on the others, in push-to-talk mode; toggle mode is optional. The macOS .fn and Globe keys cannot be bound because macOS emits no real keypress for them. On Wayland, the compositor owns the hotkey and calls yap toggle on a local socket, because Wayland blocks apps from grabbing global hotkeys.
How does yap compare with Wispr Flow?
The README's own comparison: same push-to-talk workflow, but yap runs locally by default, has no word cap and no subscription, does not send audio to a server by default, does not screenshot the active window, and is MIT open source. yap covers Mac, Windows and Linux only, with no phone app. See the open-source alternatives to Wispr Flow page for other options.
Does yap learn my vocabulary?
Yes. yap vocab add biases Whisper toward specific spellings; yap vocab fix is a guaranteed find-and-replace; and a "relearn hotkey" (default ⌘⌥L on macOS, Ctrl+Alt+L elsewhere) reads your last correction from the clipboard and stores the differences. The README also says yap watches what you dictate and adds repeated proper nouns to your glossary automatically, capped so it cannot run away, and configurable under learning.
Sources
- yap README and LICENSE: github.com/AkuchiS/Yap, read 2026-09-28. The README carries the Wispr Flow comparison table used in this page.
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 AGPL-3.0 voice typing app for Windows that ships seven local speech models including Nemotron 3.5 ASR and three Qwen3-ASR sizes, with an editable AI cleanup step through Ollama or any OpenAI-compatible endpoint.
Added September 28, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 28, 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/AkuchiS/Yap.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.