Skip to content

Latest commit

 

History

History
112 lines (88 loc) · 2.84 KB

README.md

File metadata and controls

112 lines (88 loc) · 2.84 KB
Raw

fking-sd-utils

A fking StableDiffusion utility library

A small collection of utilities to help your StableDiffusion/Dreambooth experience. This library is currently focused on assisting with captioning concept images for training in TIs, Hypernetworks or Dreambooth.

Usage

The captioning util is a hierarchical based captioner, it will apply the tags of the parent directories to each of its subdirectories and images. Each directory can contain a text file named __prompt.txt, all text in this file will be prepended to all child images, and subdirectories and their images. Using __folder__ in the __prompt.txt file will automatically use the directory's name.

Example File Tree

D:\TEST\TRAINING_DATA
│   __prompt.txt
│
├───exterior
│   │   120160.png
│   │   120161.png
│   │   120162.png
│   │   148.png
│   │   149.png
│   │   54.png
│   │   __prompt.txt
│   │
│   ├───cityscape
│   │       120228.png
│   │       120229.png
│   │       120272.png
│   │       120273.png
│   │       120274.png
│   │       120275.png
│   │       __prompt.txt
|   |
│   ├───grasslands
│   │       120166.png
│   │       120167.png
│   │       120168.png
│   │       120169.png
│   │       __prompt.txt
│   │
│   └───mountains
│           120265.png
│           120266.png
│           __prompt.txt

Run with no parameters to enable working through the UI.

py main.py

UI_Example_1

Executable

You can build your own executable of the UI using pyinstaller, after compilation the executable will be available in the dist directory.

pyinstaller fking-captioner.spec

Utility Usage

Flatten input dataset into output directory.

py main.py --no-ui -i "input_directory" -o "output_directory"

Extended Usage

You can also create text files named __special.txt in each directory, or just the root input folder is what I do, to allow you to make quick S/R tags, for example:

[
  {
    "special_tag": "__black_and_white__",
    "tags": "black and white, b&w, monochrome",
    "mode": 1
  }
]

And any image file or __prompt.txt file using __black_and_white__ will be replaced with black and white, b&w, monochrome. The modes indicate how it should handle if a parent in the hierarchy has the same special tag.

    1 - Merge tags with parent
    2 - Replace with parent
    3 - Keep current

By default, if mode is missing, it will use merge.

Final Thoughts

Don't expect things not to be broken, Python isn't my strongest language, and I am not concerned with breaking changes, I will be pushing whatever to this repo. Feel free to make any pull requests you'd like. It would be nice to have some collaboration.