amywork777 / lipflow

# Lipflow
**Wispr Flow for your lips.** Hold a key, silently mouth what you want to say, let go, and the
text shows up at your cursor in whatever app you're in. No microphone and no sound, just your webcam.
Everything runs locally on your Mac, Windows PC, or Linux. An optional LLM pass fixes the words lip reading gets wrong.
 [code] 
## Setup
Needs an Apple Silicon Mac on macOS 13 or later (macOS 26 for the Liquid Glass look), and about 2 GB of disk.
 [code] 
The first launch opens setup, which takes about 8 minutes:
1. **Permissions:** Camera, Input Monitoring and Accessibility, each asked once for "Lipflow".
 The app is a small signed program that loads this checkout in-process, so those
 permissions belong to Lipflow rather than to Python. If a switch is on in System
 Settings but setup still shows it off, click **Restart Lipflow** under Continue.
2. **Your words:** if you use Wispr Flow, it imports your dictation history (read locally, never
 uploaded) to learn your phrasing and the names you say.
3. **Practice:** you silently mouth 24 sentences (taken from your own history when available).
4. **Train:** it fine-tunes the language model on your phrasing and the lip reader on your face,
 on this Mac's GPU. Six practice sentences are held out, and the face model is only kept if it
 reads them better than the stock model. You see the before/after score.
Re-run it any time from the menu → *Set up / train on my face…*. More practice clips help more.
Everything personal (clips, phrases, trained models) lives in
`~/Library/Application Support/Lipflow/`. Delete that folder to start over.
To start Lipflow at login: System Settings → General → Login Items → add Lipflow. Logs are in
`~/Library/Logs/Lipflow.log`. Running from a terminal (`uv run lipflow`) also works, but then
macOS asks for permissions in the terminal's name.
### Windows
Needs Windows 10 or 11 (64-bit) and about 3 GB of disk. In PowerShell:
 [code] 
Then open **Lipflow** from the Start menu. It lives in the system tray (the pink mouth by the
clock; click ^ if it's hidden). The first launch opens the same setup as on a Mac, minus
permissions: your Wispr Flow words, 24 practice sentences, then training on your face.
How it differs from the Mac version:
- **Key:** hold **Right Ctrl** (change it in the tray menu → *Push-to-talk key*: Right Alt / AltGr,
 Left Alt or Right Shift also work). Double-tap for hands-free and Esc to cancel work the same.
- **Camera:** if it doesn't come on, turn on Settings → Privacy & security → Camera → *Let desktop
 apps access your camera*. Pick another camera by number in the tray menu.
- **Speed:** without an NVIDIA GPU the lip reader runs on the processor, so expect a second or two
 more per sentence than on a Mac (on an M4 Pro the encoder takes 1.7 s on CPU vs 0.16 s on the
 GPU), and training on your face takes several times longer. With an NVIDIA card, install the
 CUDA build of PyTorch: `uv pip install torch --index-url https://download.pytorch.org/whl/cu