---
title: GitHub Graphics Generator
emoji: 🎨
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.2.0
app_file: app.py
pinned: false
license: mit
---

# 🎨 GitHub Graphics Generator

Generate clean, professional graphics for GitHub repositories using a hybrid approach:
AI-generated visual backgrounds combined with deterministic text and layout rendering.

This ensures readable text, accurate stats, and consistent design.

---

## Features

### 🔹 Input Modes
- **GitHub Repository URL**
- **README / Free Text Prompt**
- **Screenshot / Image-to-Image**

### 🔹 AI Background Generation
- Text-to-Image and Image-to-Image via Hugging Face Inference API
- Style-driven prompts (Modern, Minimal, Professional)
- Custom color schemes

### 🔹 Deterministic Layout
- Project name
- Description
- Stars / forks
- Primary language
Rendered using PIL (not AI hallucination)

### 🔹 Advanced Prompt Controls
- Custom prompt override
- Negative prompt (e.g. *blurry, watermark, distorted*)

### 🔹 Export
- High-quality PNG
- One-click download

---

## How It Works

| Component | Responsibility |
|---------|----------------|
| AI Model | Visual style & background |
| PIL | Text, stats, layout |
| GitHub API | Repository metadata |
| Gradio | Interactive UI |

This avoids common AI issues with text rendering and layout precision.

---

## Supported Models

- `stabilityai/stable-diffusion-xl-base-1.0`
- `runwayml/stable-diffusion-v1-5`

Only models compatible with the Hugging Face **Inference API** are used.

---

## Environment Variables

Optional but recommended:

```env
HUGGINGFACE_TOKEN=hf_xxxxxxxxx
Provides higher rate limits and access to gated models.

⸻

Known Limitations (Intentional)
	•	No manual step / CFG sliders (Inference API controlled)
	•	AI does not render text or logos
	•	Diffusers-only models (e.g. SDXL-Lightning) are not used

⸻

License
MIT