Refine any text with AI — anywhere on Windows — in one hotkey.
Select text in any app → press Alt + Shift + W → the text is instantly replaced with an AI-polished version. Runs silently in the system tray. No subscriptions, no accounts, bring your own free API keys.
⬇ Download for Windows (v1.0.0)
Unzip → run PromptRefiner.exe. Windows 10 / 11, no install required.
First launch downloads the local AI model (~1 GB). Or add a free cloud key for instant sub-second responses.
| Provider | Avg speed | Free tier | Get key |
|---|---|---|---|
| Cerebras | ~0.3 s | ✅ Yes | cloud.cerebras.ai → API Keys |
| Groq | ~1–3 s | ✅ Yes | console.groq.com → API Keys |
Paste in: System Tray icon → Settings → Providers
- 🔥 Global hotkey — works in Word, browser, email, Notepad, Slack, anywhere
- 🤖 Three AI backends — Local (offline, private), Groq (70B), Cerebras (fastest)
- 📚 Prompt library — switch between prompt styles with one click (Alt + Shift + E)
- ✏️ Custom prompts — create, edit, colour-code your own refinement styles
- ⌨️ Configurable hotkeys — change shortcuts to anything you like
- 🔒 Private — local model never sends data anywhere; cloud keys stay on your machine
- 🔁 Auto-fallback — cloud providers automatically fall back to local on failure
1. Select any text in any app
2. Press Alt + Shift + W
3. ⚡ pill appears while AI runs
4. Text is replaced in-place with the refined version
Press Alt + Shift + E (or tray → Prompt Library) to switch prompt styles:
| Action | How |
|---|---|
| Activate a prompt | Single click |
| Edit a prompt | Double click |
| Delete | ✕ button |
| Add new | ＋ Add button |
Included prompts: Refine Prompt, System Prompt, Formalize, Simplify, Technical Depth, Expand
| Action | Default |
|---|---|
| Refine selected text | Alt + Shift + W |
| Open Prompt Library | Alt + Shift + E |
Change in Settings → General → Hotkeys → Record
Requirements: Python 3.10+, Windows 10/11
git clone https://github.com/sprawf/PromptRefiner
cd PromptRefiner
# One-click setup (handles venv + GPU detection)
install.bat
# Or manually:
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
# GPU-accelerated local AI (pick one):
# NVIDIA:
pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121
# AMD (Vulkan):
pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/vulkan
# CPU fallback:
pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
python main.pypip install pyinstaller
pyinstaller PromptRefiner.spec
# Output: dist/PromptRefiner/PromptRefiner.exe| Layer | Technology |
|---|---|
| UI | CustomTkinter (dark, GPU-composited) |
| Local AI | llama-cpp-python + Qwen 2.5 1.5B GGUF (CUDA / Vulkan / CPU) |
| Cloud AI | Groq API, Cerebras API |
| Hotkeys | keyboard library (Windows low-level hook) |
| System tray | pystray |
| Packaging | PyInstaller (--onedir, UPX disabled) |
- Local mode: text never leaves your machine
- Cloud mode: text is sent to Groq / Cerebras (both have privacy policies)
- Config + API keys stored in
%APPDATA%\PromptRefiner\config.json— never in source - Logs in
%APPDATA%\PromptRefiner\app.log(auto-rotated, max 4 MB) - No analytics, no telemetry, no accounts
MIT — do whatever you want with it.