Open Asifzm opened 1 year ago
Can you project your information onto an image? The pose format ultimately also wouldn't have to be an image but the projected onto an image and controlnet just "connects the dots".
I collect information about this topic here: https://civitai.com/articles/2078#heading-3617 -> Q: Is it possible to use non-image inputs for control nets?
See [Double Control] What model is most needed? and adding more "hint" to training process
Hi, I am trying to train ControlNet with my own data. I use this great tutorial at blob/main/docs/train.md My data has no spatial hint, but I still want to customize SD with my own image domain and corresponding prompts. How can I do that?
Thanks!