Download Latest Version 0.6.4-Beta (The SD3-Medium Support Update) source code.tar.gz (8.3 MB)
Email in envelope

Get an email when there's a new version of StableSwarmUI

Home / 0.6.4-Beta
Name Modified Size InfoDownloads / Week
Parent folder
0.6.4-Beta (The SD3-Medium Support Update) source code.tar.gz 2024-06-15 8.3 MB
0.6.4-Beta (The SD3-Medium Support Update) source code.zip 2024-06-15 8.5 MB
README.md 2024-06-15 2.7 kB
Totals: 3 Items   16.8 MB 0

This is the final release of StableSwarmUI under Stability AI.

The original developer will be maintaining an independent copy of this repo, see the Migration Guide here

Major Updates

  • added new model downloader utility tab https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-9717490
    • it can autodownload civitai metadata, and you can also use the civitai metadata importer to load metadata onto an existing model
  • New theme "Punked", inspired by Cyberpunk/Edgerunners, created by user @SampleTexting - it includes a Comfy skin too!
  • There's now a Runpod template in the readme from @nerdylive123
  • refiner do tiling parameter for tiled upscale in the refiner, thanks to code from @bash-j
  • SD3-Medium full support!
    • including SD3 TextEncs new param to select what textencs to use on the fly. Will autodownload the textencs if you don't already have them.

Minor Updates

  • added image output webp option
  • the default image format was changed to png for new installs (previously jpg)
  • updated language files for all languages (AI translation)
  • allowed steps=0 to fully load a workflow but not actually generate anything
  • got AMD directml windows working, barely
  • when editing model metadata to add a preview image, the image metadata is now preserved, so you can drag the thumbnail into the center area to copy the parameters you used to make it
  • new user setting CenterImageAlwaysGrow for if you're generating small images on a large monitor
  • new user setting AutoSwapImagesIncludesFullView for if you like the FullView and want to have it autoupdate to new images as they come in
  • added Segment Mask Blur advanced param to control segmentation better
  • added new Segment Mask Grow option to control the mask-grow factor of <segment: usage
  • added Segment Threshold Max advanced param to further assist tight control of autosegment masks
  • Comfy self-start backends now have a default-enabled AutoRestart option, to restart the backend if it crashes
  • added ImageMetadataPerFolder option for if you want to disable the metadata file showing up in image output folders
  • added a setting in server config for cloudflared path, for convenience vs CLI args
  • version displayed in the bottom right now includes a commit date to reduce updating confusion
  • also logs now have some more info at startup for troubleshooting usage as well
  • as usual, many small improvements, bug fixes, better error checking, etc.
Source: README.md, updated 2024-06-15