AI-powered image editing with a modern glassmorphism interface
Features | Installation | Usage | Screenshots
- AI Image Editing — Transform images using natural language instructions
- AMD GPU Support — DirectML acceleration for AMD GPUs (RX 6000/7000 series)
- NVIDIA CUDA Support — Full CUDA acceleration for NVIDIA GPUs
- 12 Quick Presets — Watercolor, Winter, Anime, Oil painting and more
- Batch Generation — Create multiple variations with different seeds
- Before/After Comparison — Side-by-side view of original and result
- Favorites — Save your best generations
- Auto-save — All results saved automatically with full logging
- Smart Presets — Fast/Balanced/Quality modes for different needs
- Modern Dark UI — Glassmorphism design with smooth animations
| Component | Minimum | Recommended |
|---|---|---|
| OS | Windows 10/11 | Windows 11 |
| RAM | 16 GB | 32 GB |
| GPU | AMD RX 5000+ / NVIDIA GTX 1060+ | AMD RX 6800+ / NVIDIA RTX 3070+ |
| VRAM | 8 GB | 16 GB |
| Disk | 10 GB free | SSD recommended |
Note: GPU is required. CPU-only mode is not supported due to extremely long generation times.
# Clone the repository
git clone https://github.com/randomu3/instruct-pix2pix-studio.git
cd instruct-pix2pix-studio
# Run setup (creates venv and installs dependencies)
setup.bat
# Launch the application
run.bat# Create virtual environment
python -m venv venv_win
venv_win\Scripts\activate
# Install dependencies
pip install -r requirements-windows.txt
# Launch
python app.py- Open http://localhost:7860 in your browser
- Upload an image
- Enter an instruction in English (e.g., "make it winter with snow")
- Click "Generate"
- Wait ~20-40 seconds for the result
| Do | Don't |
|---|---|
| Use clear instructions | Vague descriptions |
| "Add sunglasses to the person" | "Make it better" |
| "Turn into watercolor painting" | "Change the style" |
| "Make the sky sunset orange" | "Different colors" |
| Parameter | Description | Recommended |
|---|---|---|
| Steps | More = better quality, slower | 20-25 |
| Image CFG | Higher = preserves more of original | 1.3-1.8 |
| Text CFG | Higher = follows prompt more closely | 7-9 |
| Seed | -1 for random, or specific number for reproducibility | -1 |
Click to expand
Modern glassmorphism UI with gradient accents.
Create multiple variations at once.
System information and parameters.
instruct-pix2pix-studio/
├── app.py # Entry point
├── src/
│ ├── generator.py # Generation logic
│ ├── pipeline.py # Model loading and device detection
│ ├── ui.py # Gradio interface
│ ├── styles.py # Custom CSS
│ ├── presets.py # Prompt and settings presets
│ └── storage.py # File saving and logging
├── outputs/ # Generated images (gitignored)
│ └── favorites/ # Favorites
├── requirements-windows.txt
├── setup.bat # Windows setup script
└── run.bat # Windows launch script
pip install torch-directml- Reduce steps to 15-20
- Close other GPU-intensive applications
- Restart the application
- Ensure GPU is being used (check console output)
- AMD DirectML: ~20-40 sec per image
- NVIDIA CUDA: ~5-15 sec per image
Contributions are welcome! Please read our Contributing Guidelines before submitting a PR.
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes using Conventional Commits
- Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
MIT License — free to use, modify and distribute.
- InstructPix2Pix — Original model by Tim Brooks
- Hugging Face Diffusers — Pipeline implementation
- Gradio — Web interface framework
Made for the AI art community
If you find this project useful, please consider giving it a star!