Hold a button, speak, release — your words appear instantly wherever you're typing.
No internet required. Everything runs locally on your Mac.
Open the Terminal app (find it in Applications → Utilities → Terminal) and paste these commands:
# Download Say Stuff
git clone https://github.com/enqam/say-stuff.git ~/say-stuff
cd ~/say-stuff
# Run the installer
chmod +x install.sh
./install.shThe installer will guide you through everything. Just follow the prompts!
Once installed:
- Click anywhere you can type (a chat app, email, document, etc.)
- Hold
⌥⌘V(Option + Command + V) on your keyboard - Speak while holding the keys
- Release — your words appear!
| Keys | What it does |
|---|---|
⌥⌘V |
Record → Auto-detect language |
⌥⌘E |
Record → English only |
⌥⌘C |
Record → Russian only |
| Icon | Meaning |
|---|---|
| ⏳ Arming mic | Getting ready... |
| 🎙️ Recording | Speak now! |
| ⏳ Transcribing | Processing your speech... |
| ✅ Dictation pasted | Done! Text has been inserted |
During installation, you'll choose an AI model. Bigger models = better accuracy but need more RAM.
| Model | Size | RAM Needed | Best For |
|---|---|---|---|
large-v3-turbo |
1.6 GB | 16+ GB | Best quality — recommended for modern Macs |
medium |
1.5 GB | 8+ GB | Great balance of quality and speed |
small |
466 MB | 4+ GB | Fast, good accuracy |
base |
142 MB | 2+ GB | Very fast, basic accuracy |
-
Download a new model:
cd ~/whisper.cpp/models ./download-ggml-model.sh medium
-
Edit the config file:
open -e ~/.hammerspoon/init.lua -
Find this line and change the model name:
local model = os.getenv("HOME") .. "/whisper.cpp/models/ggml-large-v3-turbo.bin"
Change
large-v3-turboto your new model (e.g.,medium). -
Reload Hammerspoon (click its menu bar icon → Reload Config)
- Check your microphone is working in System Settings → Sound → Input
- Grant microphone access: System Settings → Privacy & Security → Microphone → Enable Hammerspoon
- Open Hammerspoon from Applications
- Grant accessibility access: System Settings → Privacy & Security → Accessibility → Enable Hammerspoon
- Reload config: Click Hammerspoon menu bar icon → Reload Config
- Make sure the window where you want text is focused (click on it) before releasing the keys
- Try a smaller model (see "Choosing a Model" above)
- Close other heavy apps to free up RAM
The installer sets up:
- Homebrew — Mac's package manager (if not already installed)
- ffmpeg — Records audio from your microphone
- Hammerspoon — Listens for your hotkey presses
- whisper.cpp — The AI that converts speech to text (runs on your Mac's GPU)
Everything runs 100% locally — your voice never leaves your computer.
To remove everything:
# Remove the script
rm -rf ~/.hammerspoon
# Remove whisper.cpp
rm -rf ~/whisper.cpp
# Remove Hammerspoon (optional)
brew uninstall --cask hammerspoon
# Remove ffmpeg (optional)
brew uninstall ffmpegMIT — free to use, modify, and share.