Getting Started with Saytrue

Installation and Setup

Saytrue runs fully offline, without a cloud account. whisper.cpp transcribes locally (optional, for microphone input); on Macs with Apple Silicon, MLX-Whisper is an alternative. Ollama and Apple Intelligence analyse locally. OpenAI, Anthropic and Mistral are optional cloud providers.

At a Glance

  1. Install
  2. Permissions
  3. Language Model
  4. AI Provider
  5. Done

Install

Free

macOS

Apple Silicon & Intel

  • App Store (in review)
  • 🍺 brew install --cask fidpa/saytrue/saytrue
  • GitHub (DMG)

Windows

Windows 10/11 (x64)

Linux

x64 & ARM64

Pro

macOS

Apple Silicon & Intel

  • App Store (in review)

Windows

Windows 10/11 (x64)

  • Microsoft Store (in review)

ℹ️ Pro is only available through the Mac App Store and Microsoft Store, as a one-time purchase without subscription; both listings are still in review. Homebrew, Snap and the packages on GitHub contain Free.

First Launch

Choose Language

15 languages available, auto-detected from system language.

Permissions

Grant microphone access (optional) and set up hotkey Ctrl+Shift+D.
Linux: global hotkeys only work in an X11 session, not under Wayland.

Storage Location

Inputs stored locally.
App Store: select folder manually.

Set Up Language Model

whisper.cpp runs locally, without internet. Only needed if you speak or load audio files. The model is downloaded once; the app picks it by your language. The quality “Standard” is preselected; “Compact” is smaller.

German Turbo

874 MB

Optimised for German. Fastest option for German transcriptions.

🌐 German · Compact: 574 MB

Large V3 Turbo

834 MB

Universal model for 13 languages. Best balance of quality and speed.

🌐 13 languages · Compact: 547 MB

Kotoba V2

781 MB

Language-optimised for Japanese.

🌐 Japanese · Compact: 513 MB

ℹ️ Optional – can be skipped and set up later in Settings.

ℹ️ On Macs with Apple Silicon, MLX-Whisper is available as an alternative. With your explicit consent, OpenAI can also transcribe; the recording then goes to the cloud.

Choose AI Provider

Offline

Ollama

Fully offline – runs locally on your device, install via shell script.

⬇️ Setup helper on GitHub

The setup helper installs Ollama, downloads a suitable model and measures the speed. On macOS it comes as “SaytrueSetup” in the DMG next to the app (GitHub, Homebrew), on Windows in the installer and on Linux in the .deb package. It is also available on its own in the GitHub release. The store editions do not include it.

or via script in Terminal:

macOS

curl -fsSL https://raw.githubusercontent.com/fidpa/saytrue-releases/main/scripts/setup-ollama-mac.sh | bash

🪟 Windows

Invoke-WebRequest -Uri "https://raw.githubusercontent.com/fidpa/saytrue-releases/main/scripts/setup-ollama-win.ps1" -OutFile "$env:TEMP\setup-ollama-win.ps1"; & "$env:TEMP\setup-ollama-win.ps1"

🐧 Linux

curl -fsSL https://raw.githubusercontent.com/fidpa/saytrue-releases/main/scripts/setup-ollama-linux.sh | bash
ollama.com →
Cloud

OpenAI

Cloud-based – fast, API key required.

Create API Key →
Cloud

Anthropic

Cloud-based – Claude models, API key required.

Create API Key →
Cloud

Mistral

Cloud-based, data stored in the EU by default (Paris), API key required.

Create API Key →
On-Device

Apple Intelligence

Runs directly on your device – no API key, no account, no setup. Requires macOS 26+ and Apple Silicon (M1 or later) with Apple Intelligence enabled.

🔐 The app stores API keys encrypted in the operating system's credential store; on Linux, this requires gnome-keyring or KWallet. Cloud usage requires GDPR consent in the app.

Need help?

Questions about setup or a technical issue?
Write to me – I'm happy to help.

Ready to go!

Press hotkey Ctrl+Shift+D or start your input in the app.