TODO - Interpause/auto-sd-krita GitHub Wiki
If you can solve any of these, feel free to make a pull request.
Documentation
Feature Videos
The Install Video Guide (especially now that it should be even easier)
Short, under 5 mins videos demonstrating each feature.
- Inpainting
- Selecting region vs using whole image
- Sketching + inpainting
- Rerolling features
- Removing boundaries (i.e. from img2img, outpainting or txt2img)
- Photoshopping in characters
- Txt2Img
- base size/max size system for highres fix
- AUTO's highres fix
- Removing background from txt2img to use as stamps
- Img2Img
- Sketching
- Creating variants of a good image
- Selecting a polygon for img2img followed by inpainting to clean up
- Outpainting
- Sketching + outpainting
- Txt2Img/Img2Img + outpainting
- Outpainting MK2
- Upscaling
- SD upscale
- Conventional upscale
Maintenance
- Fix bugs due to upstream internal API changes.
- Improve & document the API.
Upstream
- Hypernetwork selection.
- Fix new samplers that need configuration.
- Expose sampler parameters with explanation.
- Expose more config options in the GUI.
- Seed variation
- scripts
Features
- Outpainting, of which possibilities are:
- Dedicated outpainting tab
- Using Sketch + Inpaint as Outpaint technique
- Use outpainting Mk 2.
- Save prompts used per Krita document.
- Make it more convenient to send images to/from webUI.
- Show request queue & allow request interruption.
- More user feedback.
- Figure out APIs for live progress feedback.
- API to feedback token limit.
- Use alpha channel instead of white channel for masking
- The pipeline actually treats the luminance mask as alpha during blending. take advantage of this for finer blending control
- investigate using the scribbled corrections itself as the mask instead of needing a separate inpaint layer
Bugs
Unrelated
- Hmmm tag auto-complete... How would I do that in PyQt? https://github.com/DominikDoom/a1111-sd-webui-tagcomplete