- Support SD1
- Fixes when loading gguf-models
- Added BrainBlip interrogator
- StabilityMatrix compatibility
- Added Llama 3.2 button
- Support for Flux GGUF unet models
- RemBG
- Img2STL
- Florence as default Interrogator
- Pixart and Flux oh my!
- MergeMaker
- Merge-recipes for checkpoints
- Bugfixes
- Now works with the SD3 LARGE model.
- OBP Update!!!
- SD3 BBY
- Support animated thumbnails
- New LoRA selection
- Updated OBP
- New model selection with model previews!
- Minor Fixes
- Added HyperPrompt to styles / OBP
- Added handy hints!
- Fixed missing clip skip in some performances
- Updated Comfy Version
- Tidied Code
- Added CLipSkip
- Added search pipeline, Type "search:" in prompt to get todays images
- Added LayerDiffuse to create transparent images with an optional background image
Powerup Tab
- Artify Update: NMake sure you check the style selection for a TON of new styles to try
- OBP Update
- Added llyasviel's prompt tokenizer model for improving prompts, This can be used by adding the
Flufferizer
style
- Moved upscale to its own pipeline
- Various Bug Fixes
- Updated to latest comfy version
- Now supports Playground 2.5 model!
- Various Bugfixes
- Adds the ability to upgrade to a newer torch and xformers, just create a blank
reinstall
file in the RF directory
- Added presets / custom preset for OBP
- Various Bugfixes
- The OBP Update
- added "the tokinator" mode
- Fintetunes on autonegative prompt: -- Landscape removed
- Added list of artist shorthands (like van gogh)
- Added secret sauce mode (mixing OBP + SDXL styles). Happens more on lower insanity levels
- Added more consistency on lower insanity levels
- EVO with OPB variant works better:
- Small prompts now work
- (almost) always changes something
- Fix for line ends creeping into EVO + OPB Variant
- Added loads of new styles into the lists
- Added loads of more fictional characters into the lists (finally, harley quinn is in :p)
- Automatic Negative prompt, save yourself the heartache and hassle of writing negative prompts!
- Fixed some dodgy regex to support wildcards with
-
in the name
- Added better error descriptions for the civit api
- Automatically grabs lora keywords from civit on startup
- Advance prompt editing - See Wiki
- Moved Evolve to the
Powerup
tab to be cleaner - Bufixes
- Changed so loras with triggerwords textfiles dont automatically get added, instead displays trigger words to add manually
- Fixed issue with controlnet caching
- Custom resolutions can now be saved directly, just select custom from the resolution dropdown (Lavorther)
- re-enabled comfy messages
- Make Custom option for Performance Mode copy last selected mode's setings (Lavorther)
- Fixed an issue with wildcards and strengths getting stuck in Catastrophic backtracking
- Changed a lot of os.path to pathlibs
- Worked on backend code for future extensibility
- Added support for --auth
- Fixed an os.path issue by replaceing with pathlib
- Now displays 🗒️ at the end of the loraname if it has a keyword .txt file
- Prompt Bracketing Now works ie
[cat|dog]
- Updated comfy version
- More Tidying and tweaks
- Autoselect LCM lora if Lcm is selected from the quality or sampler dropdowns
- Code Tidying and tweaks
- Added Evolve which generates variations on your generated picture.
- Fixed wild imports
- Added instant One Button Prompt
- Added keybind of
ctrl+shift
to toggle `hurt me plenty`` mode - Code Cleanup
- Fixed Issue with dropdown box's being case sensitive when sorting
- Adds token count (thanks Lavorther)
- Fixed lora keywords only working the first time the lora is loaded
- Use OneButtonPrompt with infinite mode, simply put a tick in
BYPASS SAFETY PROTOCOLS
- Inpainting Update: Adds inpainting to the powerup tab
- Now comes with a built in clip interrogator, just drag your image onto the main image to generate the prompt
- Π Time - Click the small Π symbol to get a fullscreen image of your last generation great for slideshows
- updated comfy version, NEW SAMPLER TIME
- Added wildcard helper, just start typing __ and your wildcards will display
- Added slideshow
- Fixed old bug of switching models when the stop button is pressed (old code from OG-F)
- Old experimental lcm-pipeline removed
- Generate forever if Image Number is set to 0
- Updated comfy version to latest
- Nested wildcards now supported
- Random styles now correctly applying to each image
- Gradio Rollback to v3 until v4 is fixed
- WildCard Fixes
- Automatcially downloads LCM Models
- Now checks subdirectories for models
- Gradio update to V4
- New Random Style Selection
- Adding one button prompt overrides in wildcards now
- Added wildcards official
- Other stuff i'll have to get arljen to explain
- Limit seeds to 32 bits
- Sort model dropdowns
- Use caches better
- removed groop and faceswap as it was causing dependency issues on some systems
- Changed minimum cfg scale to 0
- Updated to latest comfy and diffusers (Now supports LCM Loras)
- You NEEED to set the custom settings to use lcm and sgm_sampler, steps of 4 and REALLY low config (between 0 and 1.5)
- Facewapping
- Groop
- Updated Comfy Version
- Different pipelines supporting lcm and sdxl/ssd
- Let async_worker handle model preload
- Lots of small fixes
- fixed metadata bug when stopped
- Fixed small issue with metadata not updating
- Added Metadata Viewer for Gallery items (Viewable in
Info
Tab) - Refresh Files now also reloads your
styles.csv
file
- Automatically download 4xUltrasharp Upscaler
- Added the ability to upscale images wth upscaler of your choosing
- Changed Powerup Settings so if there is a missing key from defaults it adds it to your custome settings.
- Refactored backend code to allow for future pipeline changes
- Automatically read triggerwords from <lora_filename>.txt
- Updated Comfy Version
- Added support for SSD-1B Models
- Removed ref redundant code.
- Update Comfy version and fix changes :D
- Improved image2image and allowed settings to be changed when "custom" is selected form the PowerUp Tab.
- Added the basics for image 2 image
- Renamed Controlnet to PowerUp
- Now uses
powerup.json
as default
- Wildcards can now use subdirectories
- Fixed issue where if you placed 2 placeholders with the same name, you got the same results, a new one is now chosen
- Updated status to show model loading / vae decoding
- Update to one button prompt (provided by Alrjen)
- Custom Controlnet Modes
- minor bugfixes
- moved the controlnet tab to its own ui file.
- Added sketch controlnet!
- Updated gradio version
- Added recolour controlnet!
- Restored gallery preview on all images
- renamed more variables to make sense
- bugfixes
- Added all the settings/customization to their own
settings
folder NOTE: you will need to move / copy your settings into the new directory - Bugfix where clicking stop mid-generation stopped working
- code cleanup and some renaming
- inference model speed up
- now only shows gallery when needed
- removed metadata toggle, it will now always save metadata
- save your own custom performances
- tidied ui
- fix crash when failing to load loras
- hide all but one lora dropdown showing "None"
- change fooocusversion.py to version.py for easier updating
- Moved controlnet to its own tab for easier updates
- updated gradio version
- minor wording changes
paths.json
will now be updated if there are any missing defaults paths
- Now supports controlnet
- Updated onebutton prompt so you can now add multiple random prompts by clicking the
Add To Prompt
button
- Update comfy version - Lora weights are now calculated on the gpu so should apply faster
- Bug fixes and backend updates
- changed
resolutions.csv
toresolutions.json
- updated readme
- Prompt now splits correctly using
---
- added the ability to change styles in the prompt by using style:stylename
- Added init image
- Fixed issue with wildcards if file not found.
- Fixed sorting on subfolders, so directories are displayed first
- Allowed main image window to recieve drag and drop
- Added a gallery preview underneath that will activate image window.
- Added support for subdirectories with models/loras so you can get all organised!
- showed imported image in gallery
- moved
---
split into prompt generation - correctly updates progressbar
- fixed importing of width / height
- In the json prompt, setting a seed of
-1
allows you to generate a random seed
- Render different subjects so you can process a whole list of prompts. Seperate each prompt by placing
---
on a new line
- New Beginnings. The official start of the updates!