let's go 🚀

Getting Started

From zero to your first transcription in under 5 minutes.

⚠️
Requirements
  • macOS 14.0 (Sonoma) or later
  • Apple Silicon (M1 or later) — required for on-device ML
  • 8GB RAM minimum (16GB recommended for best models)
  • ~2–4GB disk space for AI models
1

Install YapYap

Choose your preferred installation method:

⬇ Direct Download recommended

Get the latest .dmg from GitHub Releases. Drag YapYap.app to your Applications folder.

Download .dmg →
🍺 Homebrew

Install and update via Homebrew Cask.

brew install --cask yapyap
🔨 Build from Source

For developers who want to modify and run locally.

git clone https://github.com/sunboylabs/YapYap.git
cd YapYap
brew install xcodegen
xcodegen generate
make build
2

Grant Permissions

YapYap needs two permissions. On first launch it will prompt you for both.

🎤
Microphone
Required to capture your voice. macOS will show a standard permission dialog when you first try to record. Click Allow.
♿
Accessibility
Required to paste text into other apps. You need to toggle this manually in System Settings.
  1. Open System Settings → Privacy & Security → Accessibility
  2. Find YapYap in the list and toggle it ON
⚠️ If you build from source, you must re-grant this permission after every build. The code signature changes, and macOS silently revokes the grant.
3

Download Your First Model

YapYap ships without models — you choose what to download. On first launch, the onboarding flow walks you through this. Or: Settings → Models.

Recommended starter setup
STT: Parakeet TDT v3 ~600MB
Fastest, runs on Neural Engine (doesn't use your RAM budget). Good for English and 5 other languages.
LLM: Gemma 3 4B ~3.0GB
Best cleanup quality in the medium tier. 140+ languages. Recommended for 16GB+ Macs.
On 8GB RAM: use Qwen 2.5 1.5B (~1GB) instead.

Models download from HuggingFace and are stored in ~/Library/Application Support/YapYap/models/. After downloading, no internet is needed.

4

Make Your First Recording

⌥ Option + Space Push-to-Talk (default)
Hold down, speak, release. Text appears where your cursor is.
⌥ Option + ⇧ Shift + Space Hands-Free Mode
Toggle recording on/off. Auto-stops when it detects silence.

The floating bar (the little creature) appears while you're recording. The creature animates differently during different states — sleeping in the menu bar, attentive while listening, spinning while transcribing.

5

Explore the Modes

🎙️ Dictation
The default — speak naturally and get clean, formatted text pasted wherever your cursor is.
✏️ Command Mode
Highlight text in any app, press the hotkey, say a command ("make this more formal") — AI rewrites it in place.
📋 History
Browse and copy all past transcriptions from the popover or history window.
⚙️ Settings
Click the creature in the menu bar → Settings. Or use the keyboard shortcut shown in the popover.
🧠
Choose your models
Learn which speech and language model is best for your machine and use case.
Models guide →