Recent Updates via Github:
Updates include better folder management and controlling which images are sorted based on a flagging system. If the images are sorted, a flag is created and stored in the folder. Then when you run the program again, it will skip those folders and only organize files in the top folder or subfolders (if you want). Also, the metadata files are disabled by default unless you want them (just set it in config). This was mainly a quality of life update to the program.
New project:
An application based program (possible cloud future integration) which serves as a manual + future CLIP/BLIP / API based tagging program to assist in tagging images. It is a work in progress, but making progress fairly quickly.
Here is a screenshot of the program. I haven't added all the tools I've designed nor have I updated the theme...but all the widgets are draggable droppable so you can design the layout like you want. It also supports user created plugins. This isn't the final design but this is a day's work.
Images are shown are going to be for my 1800s dress oil painting Lora that I may or may not get to complete.
It is designed via python and the gui is PySide6.
I'll be sure to add this to github when finished, and it will probably be my best program to date. Designed for me in mind, but feel free to download it yourself when it's finished / in beta :)
This program is designed as a tool that you can use in your regular Lora training programs, and is not meant to replace those programs, only help you to tag your images correctly with your prompts and trigger words. Helps you to visually see and filter tags (prompts) for your image sets.
The tag filter system lets you, the user, assign the category of the image, eg. PG, or PG-13 or a custom filter setting, and then when API functionality is added so you can use AI to help with prompts or to use other programs to help with tagging for prompts, I'm hoping that the filter rating will help find the right tags for the image.
This could end up being a fully featured premium program that I upload to a website or allow people to buy so I may not end up putting it on github...but this is a program I'm hoping to get somewhat polished, at least for a beta release. If not, at least it will be a program I can use for myself and when it eventually finishes be able to release it.
New Lora
The images in the application are going to be part of an 1800s themed dress Lora which I hope will work pretty good at capturing how the dresses are worn, the materials used, the way it functions under different lighting, and it's an oil based painting style in all the images, so I'm hoping it will capture the unique brush strokes that are present in the images and hope to pass on through lora training. The oil based paint style is a broader lora that I won't have time to finish this month, so it will probably be tabled for much later in the year before it gets done. This side lora for dresses has about 290 Dall-e-3 images that will be trained.
The broader lora that has the painting style but broad topics (not 1800s) I hope to be a great painting tool to make any model paint in beautiful oil-based paint. Building off my success of my first JP-drawn-style loras.