EDIT: V2 is out! now with a video to show you around
This is a robust dataset creator for creating models and loras!
Iv built it as cleanly as I could, including the ability to toggle on "Save images" and "Save Captions"
The caption feature is experimental. I suggest only toggling these on when you are ready and happy with your images to export. Simply toggle on the settings and queue the workflow again. With all settings being fixed, it should skip all the rendering and go right to saving the files.
All files are saved in the output folder in Comfy to a "LoraData" folder. Inside that folder, you will see P1-25 folders with all your images and text files.
please note that as it stands, the workflow only saves 4 Txt files, even if you have a batch of less than 4. I have not found a way to make the amount of text files saved variable. If anyone has suggestions please message me here or on Discord about it and I can make revisions and upload a V2!
This tool is really fun for exploring a bunch of styles/subjects all at the same time!
Iv attached Global sampler settings in the bottom left, as well as global high-res fix settings. The workflow also includes an IP adapter and 3 loras, if you choose to use them.