title	colorFrom	colorTo	sdk	sdk_version	app_file	pinned
PersonaGen	blue	green	gradio	5.44.1	app.py	true

PersonaGen – AI-Generated Portraits with Stable Diffusion

---

📌 Overview

PersonaGen is a generative AI project that creates synthetic human portraits using Stable Diffusion and the 🤗 Hugging Face Diffusers library.

The system takes a text prompt as input and produces high-quality faces that can be customised by age, gender, expression, and artistic style.

👉 Try the live demo on Hugging Face Spaces: PersonaGen Demo

---

🎯 Features

• Generate photorealistic or stylised portraits from text prompts
• Control over age, gender, mood, and style
• Artistic filters: watercolour, comic, cyberpunk, oil painting
• Deployed with Gradio on Hugging Face Spaces

---

🛠️ Tech Stack

• Python 3.10+ – primary language for AI development and web app.
• PyTorch – tensor computations, model execution on CPU/GPU. (pytorch.org)
• Hugging Face Diffusers – Stable Diffusion pipelines, model management, inference. (github.com/huggingface/diffusers)
• Gradio – easy-to-use UI for running ML models in a web browser. (gradio.app)
• PEFT (Parameter-Efficient Fine-Tuning) – apply LoRA weights to enhance models with minimal compute. (huggingface.co/docs/peft)
• LoRA Weights for Stable Diffusion – optional fine-tuning for realistic facial features.
• Additional Python Libraries:
  • numpy – numerical operations.
  • Pillow – image creation and manipulation.
  • torchvision – image preprocessing and utilities.
  • ipywidgets – interactive notebook controls (optional for demos).

---

🗂️ Repository Structure

PersonaGen/
├── app.py                  # Gradio interface (main entry point)
├── requirements.txt        # Dependencies for Hugging Face Space
├── README.md               # Project documentation
├── .gitignore
├── examples/               # Pre-generated demo portraits
│   ├── woman_blonde.png
│   ├── smiling_kid.png
    └── young_male.png

---

⚡ Quick Start

Follow these steps to run PersonaGen locally or on Hugging Face Spaces:

1. Clone the repository

git clone https://github.com/yourusername/PersonaGen.git
cd PersonaGen

2. Install dependencies

pip install -r requirements.txt

3. Run the app

python app.py

4. Open the app in your browser

• Visit the local URL shown in the console (usually http://127.0.0.1:7860).

• Or use the Share link provided by Gradio to access the app remotely.

5. Optional: Enable LoRA for enhanced faces

• Install PEFT if not already installed:

pip install peft

• Check the "Apply Face LoRA" box in the interface to use the LoRA weights.

---

🚀 Installation

Clone the repository and install dependencies:

git clone https://github.com/yourusername/PersonaGen.git
cd PersonaGen
pip install -r requirements.txt

Or using conda:

conda create -n personagen python=3.10 -y
conda activate personagen
pip install -r requirements.txt

Run the Gradio app locally:

python app.py

Using LoRA (optional)

Ensure the peft library is installed (pip install peft). Enable the "Apply Face LoRA" checkbox in the app for enhanced facial features.

---

🎨 Usage

1. Launch the app with:

python app.py

2. Enter a text prompt in the Gradio interface. Example prompts:

• Modern portrait of woman with blonde hair, smiling, realistic
• Stylised portrait of a young boy smiling, beach background, watercolor style
• Portrait of a young male character, smiling, vibrant colours

3. Adjust settings:

• Steps: number of diffusion steps (default: 25)
• Style: choose from photorealistic, watercolor, comic, cyberpunk, oil painting
• LoRA: check the box to enhance facial features (requires peft)

4. Generated images are displayed in the web UI and can be downloaded by clicking the image download icon.

---

🖼️ Example Portraits

Prompt	Result with LoRA	Result without LoRA
Modern portrait of woman with blonde hair, smiling, realistic
Stylised portrait of a young boy smiling, beach background, watercolor

Caption: Images in the "Result with LoRA" column include LoRA fine-tuning for enhanced facial details, while "Result without LoRA" shows the base Stable Diffusion output.

Notes:

• Steps = number of diffusion steps (default: 25).
• CPU runtime varies depending on hardware (around 451.4s).
• LoRA enhancement requires peft and is optional.

This version is compact, public-repo friendly, and highlights LoRA, CPU optimization, and basic usage instructions.

---

🧩 References (Models & Weights)

• Stable Diffusion v1.5 – base text-to-image model
  https://huggingface.co/runwayml/stable-diffusion-v1-5

• LCM LoRA for SD v1.5 – optional LoRA weights for enhanced faces
  https://huggingface.co/latent-consistency/lcm-lora-sdv1-5