Image Generator

This image generation tool uses Stable Diffusion and Generative Adversarial Network (GAN). Take a look at the demo on Google Colab (Google Chrome is preferred).

Installation

Note: Use Python 3.10 or above.

1. Install PyTorch (find a version that suits you most).

2. Clone this repository and change directory:

   git clone https://github.com/anson416/image-generator.git
   cd image-generator

3. Download Real-ESRGAN models (realesr-general-x4v3.pth , realesr-general-wdn-x4v3.pth , RealESRGAN_x4plus.pth ) to ./imgen/gan/realesrgan/weights (backup).

4. Install imgen:

   python -m pip install -e .

Usage

Stable Diffusion

Positive Prompt

Suggested prefix: "(((masterpiece))), (((best quality))), ((ultra-detailed)), ((8k))"

If you want to prevent NSFW content, you can add "(family friendly:0.85)" after the above prefix.

Negative Prompt

Suggested prefix: "lowres, worst quality, low quality, standard quality, error, jpeg artifacts, blurry, username, signature, watermark, text"

If you want to prevent NSFW content, you can insert "((nsfw)), ((nude))" before the above prefix.

Materials

1. Prompt Templates for Stable Diffusion
2. The Code of Quintessence

Acknowledgement

• BasicSR: An open-source image and video restoration toolbox based on PyTorch.
• Real-ESRGAN: A practical algorithm for general image/video restoration.