feature request, image-based concept slider training #32
Description
hi. highly appreciate your work, was attempting myself to extend the UI by not4talent https://ko-fi.com/s/2fe3a2d863 to allow for image based flux concept slider lora training (the ui supports sd1 & sdxl), based on Rohit Gandikota's flux script https://github.com/rohitgandikota/sliders myself but having very limited coding skills, this never went anywhere. (I know AI toolkit supports text based slider training, but editing 2 states of an image is so much simpler)
Having seen you already support training by pairs for qwen image edit training, would it be possible for you to implement image pairs based concept slider training for flux, qwen, z-image turbo, other models? I'd really really appreciate if you could add option to train sliders with 2-3 pairs of images to your tool, specially for flux, qwen and zit. thanks a lot in advance, and thanks a lot for your awesome tool. (edit, i additionally requested this in Ostris' discord)