Support

Install

Install SayType with Homebrew:

brew install --cask frugoman/tap/saytype

To update to the latest version:

brew upgrade --cask saytype

Getting started

Open SayType and follow the setup window. It asks for Microphone and Accessibility access and downloads the speech model (about 630 MB, one time). Then click into any text field and talk. When you pause, the text is typed.

SayType isn't in the Accessibility list

Open System Settings → Privacy & Security → Accessibility, click + below the list, choose Applications → SayType, and turn it on.

It doesn't type in a particular app

Some apps don't tell macOS where their text fields are. In SayType's Settings → General, add the app to Always listen in these apps.

It gets a word wrong

Open Settings → Vocabulary, add the word, and click Teach. Say the word three times, and SayType learns how it's misheard and fixes it from then on.

Shortcuts

Listening modes

In Settings, choose how SayType listens:

• Hands-free (the default). It listens while a text field has focus and types when you pause.
• Push-to-talk. Hold ⌃⌥Space while you talk and release to type. Good in noisy rooms or when you don't want it listening on its own.
• Toggle. Press the shortcut to start, press it again to stop.

A small recording pill shows when SayType is listening. You can also turn on start and stop sounds.

Voice commands

Say these on their own, as a separate phrase:

• "new line" and "new paragraph"
• "scratch that" deletes what was just typed
• "undo that"
• "select all"
• "press enter"

You can also turn on spoken punctuation, so "comma", "question mark" and "open paren" type the symbol.

Snippets

In Settings, add a phrase and the text it should type. Say the phrase and the whole block is typed. In the text you can use {date}, {time} and {clipboard}, which are filled in when it's typed.

Per-app profiles

A profile changes how SayType behaves in one app. You can set the pause length, language, tone (casual, professional, code or literal), whether to remove the trailing period, and extra vocabulary. SayType suggests profiles for apps like Slack, Mail, Terminal, Xcode, VS Code and Messages.

Transcript history

Open Settings → History to search what you dictated, copy an item, or type it again. The menu bar shows your recent items. SayType keeps the last 10 items by default; change that with Items to keep (lowering it deletes the older ones). History is stored as text on your Mac, never as audio. You can turn it off or clear it in the same place.

Set up AI cleanup

AI cleanup is off by default. It removes filler words and fixes punctuation, and can also polish or translate what you say. Turn it on in Settings and choose a backend. Nothing leaves your Mac either way.

Apple's on-device model. On macOS 26 or later, pick Apple's model. There is nothing to set up.

A local server. Any server with an OpenAI-compatible API works, including Ollama, LM Studio, llama.cpp and mlx_lm.server. With Ollama, install a model, then start the server:

ollama serve

In SayType's AI settings, set the server address to http://127.0.0.1:11434 and choose the model.

Edit text by voice

Select some text, press ⌃⌥E and say what you want: "make this shorter", "translate to Spanish" or "fix the grammar". The selection is replaced. This needs an AI backend, set up as above.

Languages and translation

SayType understands about 30 languages. Choose one in Settings, or set it per app in a profile. There are two speech engines: Whisper, and Parakeet v3 (25 European languages, very fast; it downloads its own model). To get English out of another language, turn on Translate to English (Whisper). To get another language out, set an Output language, which uses the AI backend.

Transcribe a file or record a meeting

From the SayType menu, choose Transcribe a file… and pick any audio or video file. You can export the result as .txt, .srt, .vtt or .md.

Choose Record a meeting to capture your Mac's audio and your microphone. The transcript labels who spoke as You or Others, and the AI backend can add a summary with action items. Meeting recording needs one more permission: open System Settings → Privacy & Security → Screen & System Audio Recording and turn on SayType. Dictation doesn't need it. The audio isn't saved, only the transcript if you save it.

Command line and automation

Homebrew installs a saytype command:

saytype toggle
saytype speak "Build finished"
saytype transcribe ~/Recordings/interview.m4a
saytype last | pbcopy
saytype help

saytype status prints one word: listening, idle, off, loading, speaking or recording.

The same actions work as saytype:// links, for example open -g "saytype://dictation/toggle". In the Shortcuts app, search for SayType to find actions that toggle, start and stop dictation, speak text, get the last transcript and transcribe an audio file. There are also script commands for Raycast. See the integrations guide.

Is it really free?

Yes. There's no trial, account or paid version, and the code is open source (MIT). If SayType is useful to you, you can buy me a coffee.

Contact

Report a problem or ask a question by opening an issue on GitHub.