Screenshot AI Answer Tool
A desktop application that captures any part of your screen and provides AI-powered answers using either OpenAI's GPT-4 Vision or local Ollama models.
-
Always On Top Window
The app window stays above all others usingattributes("-topmost", True)so it’s always accessible. -
Two AI Providers
- OpenAI GPT-4 Vision — Cloud-based, advanced visual understanding.
- Ollama — Local AI models for privacy and offline usage.
-
Enhanced UI
- Emoji-based status indicators with color coding.
- Scrollable text area for long AI responses.
- Organized settings panel for provider and model selection.
-
Keyboard Shortcut
Press F10 to capture and process instantly.
Install Python 3.9+ and the dependencies:
pip install -r requirements.txtrequirements.txt
openai>=1.0.0
pyautogui>=0.9.54
pynput>=1.7.6
Pillow>=9.0.0
requests>=2.28.0
- Get your API key from platform.openai.com.
- Set it as an environment variable:
export OPENAI_API_KEY="your-api-key-here"
- In the app settings, choose OpenAI as the provider and select your model (e.g.,
gpt-4o,gpt-4o-mini).
- Install Ollama from ollama.ai.
- Pull a vision-capable model:
ollama pull llava ollama pull moondream
- Make sure Ollama is running (
ollama serve). - In the app settings, select Ollama as the provider, set the API URL (
http://localhost:11434by default), and choose your model.
- Run the app:
python main.py
- Configure provider in settings (⚙️).
- Select Area → drag to capture part of the screen.
- Press F10 to capture and process.
- AI-generated answer appears in the text box and is copied to clipboard.
- Cloud processing with high accuracy.
- Supports multiple GPT-4 variants.
- Fast and reliable.
- 100% local processing — privacy-first.
- Works offline after model download.
- No per-request API cost.
- Thread-Safe UI Updates: Uses
root.after()to avoid crashes. - Resource Management: Cleans up listeners and temporary files.
- Better Error Handling: Graceful fallback if provider fails.
- Organized Code: Modular functions and improved readability.
- UI Enhancements:
- Status indicators
- Scrollable output
- Resizable window
- Screenshots saved locally in the
ss/directory. - No unauthorized uploads — only sent to your selected provider.
- Environment variables protect API keys.
- App cleans up temporary resources on exit.
This project is licensed under the MIT License. See LICENSE for details.
- Add Whisper integration for audio-based Q&A.
- Add history panel for previous answers.
- Improve multi-monitor capture support.
- Dark mode UI.
Created by Halil Uğur.
