Skip to content
This repository has been archived by the owner on Jul 22, 2024. It is now read-only.

Latest commit

 

History

History
89 lines (73 loc) · 2.83 KB

TODO.md

File metadata and controls

89 lines (73 loc) · 2.83 KB
Raw

TODO

Issues

Stuff to be fixed, in no particular order...

  • Misterious Extra network corruptions
  • Unet type mismatch/autocast issues on some platforms

Features

Stuff to be added, in no particular order...

  • Diffusers:
    • Add SD-XL Lora
    • Fix SD-XL Img2img/Inpaint
    • Add VAE direct load from safetensors
    • Fix Kandinsky 2.2 model
    • Fix DeepFloyd IF model
    • Redo Prompt parser for diffusers
    • Add unCLIP model
  • Technical debt:
    • Port A1111 stuff
    • Port p.all_hr_prompts
    • Import core repos to reduce dependencies
    • Update gradio
  • Non-technical:
    • Create additional themes
    • Update Wiki
    • Get more high-quality upscalers
    • Rename repo: automatic -> sdnext
    • Localization
  • New Minor
    • Prompt padding for positive/negative
    • XYZ grid upscalers
    • Built-in motd-style notifications
    • Docker PR
  • New Major
    • Style editor (use json format instead of csv)
    • Profile manager (for config.json and ui-config.json)
    • Multi-user support
    • Image phash and hdash using imagehash
    • Model merge using git-rebasin
    • Enable refiner-style workflow for ldm backend
    • Add sgm backend
    • Cache models in VRAM
    • Train:
      • Use interrogator
      • Use rembg
      • Templates for SD-XL training
      • Lora train UI
  • Redesign
    • New UI
    • New inpainting canvas controls (move from backend to purely frontend)
    • New image browser (move from backend to purely frontend)
    • New extra networks (move from backend to purely frontend)
    • Change workflows from static/legacy to steps-based

Investigate

Stuff to be investigated...

Merge PRs

Pick & merge PRs from main repo...

Integration

Tech that can be integrated as part of the core workflow...

Random