Happy holidays everyone -

Development is still progressing towards the next update. I have created a Milestone in order to help track progress for the remaining tasks. You can take a look at that here on Github.

As of this writing I am 68% through the milestone. This number may change as I add more tasks. It is difficult to put a time estimate on the remaining work - the update is huge and will require testing and several iterations before all of the bugs are worked out, so when it does go live I will be releasing it under a separate release branch.

As a reminder, I will never charge for AI Runner upgrades or updates. v2.3.0 will be made available to anyone who has purchased the app.

Here are just some of the changes you can expect in the upcoming release:

• LLM support Ability to generate prompts from existing images, use chatbots, and more
• Improved canvas with support for dragging, multiple images per layer zoom
• Model Manager which allows for easy management of models
• CivitAI model downloader to automatically download models from CivitAI by pasting a URL
• Layer merging
• Clip skip generate option added
• Ability to delete pixels outside of the active grid area by pressing shift+delete
• Splitter widgets added for better panel management
• Easy open prompt builder button added to generator form
• Save prompts button added to generator form
• Previously active tabs are now remembered and reactivated between sessions
• Second seed for prompt builder and other interface related calls to random
• Styles moved into a QSS stylesheet and adds a style sheet builder tool
• Black image with “NSFW” in center is displayed when a NSFW image is generated and the NSFW filter is enabled
• Models and embeds directories are now scanned on load and the DB is updated with new data automatically
• Settings files are now stored in SQLite DB
• Templates are now properly set up with pyqt designer. Buttons and other actions are linked within designer and handled with widget controller files. Widgets follow a consistent pattern and are easier to develop.
• Filters have been moved to the database, making them easier to maintain. Additionally they may now be combined dynamically
• Automatically apply filters to generated images
• Combined txt2img / img2img tab. Enable input image to use img2img
• Input Image image tool which allows for improved interaction with input images (for img2img, pix2pix, depth2img etc)
• Controlnet tool which allows for improved interaction with ControlNet. Allows for the ability to import and export (as well as view) controlnet masks.
• Pipeline loading Pipelines are now loaded dynamically by using data from the DB
• Error handling and messaging improvements
• Improvements to batch generation
• improvements to socket server and first steps towards fully supporting the engine via socket server
• Improved settings manager signal handling
• Improved tracking of generator tabs in settings manager
• Various bug fixes
• Moves layers to database - images are now reloaded from DB on session reload
• Moves active grid area settings to DB
• Allows active grid area border and fill to be modified
• Allows active grid area to be disabled

There are many more additional features and bug fixes which I will outline in full when the release eventually goes live.

Until then, enjoy AI Runner and be sure to join the community here or on Discord in order get help or give suggestions for new features.