SD-Script

Small and simple Python script for playing with Stable Diffusion models.

Requirements

SD-Script has been tested with Python 3.10 and PyTorch 1.12.

CUDA

For CUDA you need to install PyTorch CUDA variant from https://pytorch.org/.

Usage

Grab requirements by running command pip install -r requirements.txt or pip install --user -r requirements.txt.

Run the script with Python by running command python sdscript.py while in same directory (assuming you only have single version of Python installed).

English commands and default values

• .read - Reads a file for inputs, each newline being a separate input. Example: .read sample.txt
  • Additionally, you can load an image file (.png, .jpg, .jpeg) to use as reference image for stable diffusion.
• .multiline - Multiline input, press Ctrl + Z and Enter when done.
• .stop, .exit, .quit - Exit from SD-Script.
• .count - (1) Amount of images generated for same input. Example: .count 3
• .help, .? - Prints help text.
• .inferences - (50) Amount of inferences per image. Example: .inferences 100
• .width - (512) Width of the generated image, converted to multiples of 64. Example: .width 1024
• .height - (512) Height of the generated image, converted to multiples of 64. Example: .height 1024
• .seed - (-1) Seed for random number generator, for reproducible results. Example: .seed 1234
• .reset - Clears set reference image.
• .strength - (0.75) Strength of the referenced image. Example: .strength 0.5
• .guidance - (7.5) How much the model should follow the prompt (higher = more). Example: .guidance 8.5
• .magic - (False) Toggles MagicPrompt model to add detail to your prompts. Example: .magic
• .magiccount - (1) Amount of MagicPrompt prompts to use. Example: .magiccount 3
• .nsfw - (False) Toggles whether NSFW output is allowed. Example: .nsfw