SD GUI 1.7.2 Changelog:

- Improved prompt syntax converter (converts A1111 or pre-1.7 weighting to new syntax)

- VAE Model selection now also lists .pt files (same format as ckpt, just different extension)

- Fixed init images being in random order after importing

- Fixed idle CPU usage

- Fixed an issue with having the same model filename in multiple model folders

- Fixed issue where k_diffusion module was not imported properly

- If a config YAML exists next to a model file it will be used instead of the default (e.g. model.ckpt.yaml)

SD GUI 1.7.1 Changelog:

- Added DPM++2 and DPM++2A samplers

- Added advanced syntax for step count (run same image with different step counts)

- New prompt attention syntax - Use (text)++ instead of ((text)), etc.

- Added automatic conversion of old attention syntax to new syntax

- Security: Added pickle malware scanner (checks model files automatically)

- Minor UI improvements, DreamBooth now has a dedicated button

- Fixed issues with pytorch not working correctly (No CUDA GPU detected)

- Fixed issues with post-processing after generation

- Fixed negative prompt input being visible when using Low Memory Mode

- Fixed certain sliders loading invalid values after restarting the program

- Reduced installation size and optimized code

SD GUI 1.7.0 Changelog:

- Added support for running multiple init images at once, e.g. for animation frames

- Added support for wildcards (insert words/phrases into prompt dynamically)

- Added "Hi-res Fix" which allows better results at higher (>512px) resolutions

- Added support for loading custom VAE models (can improve image quality)

- Added support for running upscaling or face restoration manually

- Added support for all samplers when using an init image (not just DDIM)

- Added button to resize the window to fit the currently displayed image

- Added option to save output images in a subfolder per session

- Added options to use seamless mode only on one axis (horizontal or vertical)

- Added a label that displays the current image's prompt

- Added hotkeys: Quick-switch VAE, copy current image, copy image to favorites

- Added new developer tool: Open CMD in SD Conda environment

- DreamBooth training: Images can now be automatically resized if they are not 512x512

- DreamBooth training: Added slider for steps multiplier (0.5x-2x)

- GUI improvements, prompt and negative prompt are now separate

- Image viewer now also shows "actual" image resolution (for upscaled images)

- Sliders now also allow you to type a value instead of dragging the handle

- Loading image metadata now also works for images generated with automatic1111

- "Delete All Current Images" now requires a confirmation

- Improved prompt history/queue UI (show full prompt on hover, and more)

- Improved GUI rendering, should have less flickering now

- Fixed bug where model pruning would say "Failed..." even if it was successful

SD GUI 1.6.0 Changelog:

- Added Dreambooth training (24 GB VRAM required for now!)

- Added support for prompt emphasis - () for more, {} for less

- Added Model Quick Switcher: Press Ctrl+M to quickly change the current AI model

- Added model folder manager: You can now add additional model folders to load models from

- Pop-Up Image Viewer: When slideshow mode is enabled, Left/Right arrow keys change images

- Pop-Up Image Viewer: Window can now be resized/zoomed in 25% steps using mouse wheel

- Pop-Up Image Viewer: Added "Always on Top" option to keep window always on top

- Added lots of hotkeys, documented on Github Readme

- Words in prompt field can now be deleted using CTRL+BACKSPACE, like in most text editors

- Model Pruning Tool: Added option to delete input file if pruning was successful

- Fixed an issue where the Stable Diffusion process would be killed when cancelling

- Fixed prompt queue not working after running first entry

SD GUI 1.5.0 Changelog:

- SD Update: Supports exclusion words, runs on 4 GB VRAM (when no other apps are open)

- UI is now more flexible, window can be resized, prompt field is bigger and has zoomable text

- Added CodeFormer face restoration as an alternative to GFPGAN

- Updated RealESRGAN (upscaler), should now be faster with same or better quality

- Added button to delete either the current image, or all generated images

- Added separate checkboxes for including prompt/seed/scale/sampler/model in filename

- Added option to save original image in addition to the post-processed image (if postproc is enabled)

- Added option to select the CUDA device (Automatic, CPU, or a specific GPU)

- Added model merging tool

- Added model pruning tool (strip EMA data and/or convert to fp16 half-precision for 2 GB models)

- Added option to unload Stable Diffusion after each generation (like in pre-1.4.0)

- Added reliable orphan process handling (Python no longer stays in RAM if the GUI crashes)

- Image Viewer slideshow: Added short cooldown after using prev/next image buttons

- Image Viewer context menu: Added button to re-generate single image with current settings/seed

- Image Viewer Pop-up: Now borderless, 100% zoom by default, double-click for fullscreen

- Image Viewer Pop-up: Added Slideshow Mode which mirrors the regular image viewer when enabled

- Added image load form, allows you to use as init image, load settings from metadata, or copy prompt

- Images can now be loaded from clipboard, not just from files

- Prompt History: Added option to disable history, added text filter

- You can now add an entry to the prompt queue by right-clicking on its icon

- Disabled post-processing with Low Memory Mode as it was not working properly

- Prompt text in filenames now strips weighting (won't create new folders for each weight change)

- Current model name gets printed whenever Stable Diffusion is started

- Full Precision is now enabled by default on GTX 16 series cards to fix compatibility with them

- Fixed empty/invalid prompts (e.g. newlines) counting towards the target image amount

- Some fixes regarding cancelling the generation process and handling crashes