README.md

metadata

title: Sketch To Image
emoji: 😻
colorFrom: yellow
colorTo: blue
sdk: docker
pinned: false

🎨 Sketch to Image AI

An interactive web application that transforms your rough sketches into detailed, high-quality images using Stable Diffusion v1.5 and ControlNet (Scribble).

✨ Features

• Live Sketchpad: Draw directly in the browser using the integrated Gradio Sketchpad.
• Prompt-to-Image: Describe your sketch with natural language to guide the AI's generation.
• Advanced Controls: Fine-tune the output with adjustable inference steps and guidance scales.
• Optimized Performance: Uses CPU offloading and attention slicing to run efficiently.

🚀 How it Works

1. ControlNet (Scribble): Processes your hand-drawn sketch as a structural guide.
2. Stable Diffusion v1.5: Generates a realistic or artistic image based on your prompt, while adhering to the structure provided by the sketch.
3. UniPCMultistepScheduler: Ensures fast and high-quality image sampling.

🛠️ Tech Stack

• Framework: Gradio
• Model: Stable Diffusion v1.5
• Control: lllyasviel/sd-controlnet-scribble
• Library: Diffusers by Hugging Face
• Language: Python

💻 Local Setup

1. Clone the repository:

   git clone https://github.com/your-username/Sketch-Image.git
   cd Sketch-Image
   

2. Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Run the application:

   python app.py
   

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Built with ❤️ by logan-codes