search

DrCyanide / cyanic-sd-krita

A Stable Diffusion plugin for Krita, compatible with A1111 and SD.Next
68 stars 4 forks source link
generative-ai krita-plugin stable-diffusion

readme

Cyanic SD-Krita Plugin

A Krita Plugin to work with the Automatic1111 WebUI, SD.Next WebUI, and Forge WebUI

Img2Img Example

Note: This plugin is in a "Usable Alpha" state. It might crash, or might not be compatible with your server settings. Things are expected to change somewhat frequently.

Updates

Apr 11 2024 - Added support for Soft Inpainting

Mar 30 2024 - Added a max size option in Settings, for working with a large canvas without over taxing Stable Diffusion

Mar 23 2024 - Fixed some ControlNet compatibility bugs with Forge

Jan 4 2024 - Fixed UI bug with collapsible categories (shouldn't cause squishing/overlapping now)

Dec 10 2023 - Min Size added (assists with smaller details)

Nov 15 2023 - CFG Scale support added. Refiner support for A1111 added.

Nov 2 2023 - Hires Fix support added.

Oct 28 2023 - Better Inpaint Masking options (Thank you Miraihi for all your help finding bugs)

Oct 26 2023 - Lora, Textual Inversion, and Hypernetwork lists added, under Extra Networks

Oct 25 2023 - Batch controls added, Simplify UI tab added (to hide features you don't use to make it more like Fooocus)

Oct 20 2023 - RemBG support added, ControlNet Preprocessor Previews supported.

Oct 17 2023 - Upscaler support added.

Oct 5 2023 - ADetailer support added.

Oct 4 2023 - Inpainting added.

Oct 3 2023 - First iteration of ControlNet support added.

Install Instructions

  1. Ensure your server has API access. For A1111, this means adding --api to your webui-user.bat (or equivalent) COMMANDLINE_ARGS
  2. Open Kritam then go Settings > Manage Resources, then in the bottom right click the Open Resource Folder button

  1. Find the pykrita folder inside the explorer window Krita opened up
  2. There's two options for installing the extension:
    • Easy install, annoying update - Download the project and put both the folder cyanic and the file cyanic.desktop in your pykrita folder. You'll have to redownload and recopy the files to this path any time you wish to update.
    • Annoying install, easy update - Git clone the project, then create a symlink in the pykrita folder to cyanic and cyanic.desktop. Anytime you do a git pull it'll update the plugin.
      • Windows: Open Command Prompt by right clicking it and selecting "Run as Administrator..." and run these commands 
        mklink /j "path_to_your_pykrita_folder\cyanic" "path_to_your_git_pull\cyanic"
mklink "path_to_your_pykrita_folder\cyanic.desktop" "path_to_your_git_pull\cyanic.desktop"
      • Linux: Open a terminal and run these commands 
        ln -s "path_to_your_pykrita_folder/cyanic" "path_to_your_git_pull/cyanic"
ln -s "path_to_your_pykrita_folder/cyanic.desktop" "path_to_your_git_pull/cyanic.desktop"
  3. Restart Krita and go to Settings > Configure Krita..., then at the bottom left side of the window until you see Python Plugin Manager. Click that, then check the box to enable Cyanic SD Plugin for Krita
  4. Restart Krita one more time
  5. You should see a docked Cyanic SD window in Krita. If it's not showing up, go to Settings > Dockers and check the box next to Cyanic SD

Plugin Basics

NOTE: This description was written while the app was in "Usable Alpha". The design might change as the project matures.

Since I prefer to use Krita on my laptop, I wanted this plugin to be as compact as possible. An abundance of tabs makes the WebUI's too bulky, so it was replaced with a dropdown bar in this plugin.

Similar to the WebUI, there are categories that are collapsible.

The Style selection is done by selecting a checkbox next to the name of the style. Any checked style will be passed to the server to be included with the prompt/negative prompt. Clicking "Add to Prompt" will uncheck all styles and add their corresponding prompt/negative prompts to the fields where you can fine tune them.

Selecting a source image for Img2Img has a few options. Selecting "Use Layer" will narrow the selection to the bounds of the current layer, while "Use Canvas" will select the entire canvas. NOTE: "Use Layer" doesn't expose or hide any other layers - it only refers to the size image.

Inpainting masks are super simple to make - just create a new layer and draw. Any color works, as do multiple colors, because the end mask is based on the transparency of the pixels in your active layer.

Want something a little bit more like Fooocus, with fewer settings in your way? "Simplify UI" to reduce the amount of clutter.

Not yet implemented

Not planning to implement