Compatibility on webui
Hi, I was able to apply your model in Comfy UI, but I would like to ask if it is applicable in webui (automatic1111) also.
When I tried with the test image, and the model simply produced the exact same image I used as controlnet condition (masked image).
Could you tell me if I did something wrong?
I suppose it should work with webui as this is a classic controlnet, but I did not test it myself.
You used one three image provided on this page ?
It is supposed to inpaint in the completely white area and the image is the controlnet input, you should not use it in an img2img workflow, only txt2img.
Is it what you did ?
I don' see how to use this in A1111 / Forge. It says to use txt2img and denoising strenght 1.0, but the txt2img tab doesn't have denoising strength settings as far as I can see.
Hi, Can i use it in Comfyui , i try to use your model with == Inpaint preprocessor node and Apply Controlnet node== but it s shown an error like this == > Error occurred when executing KSampler:
'NoneType' object has no attribute 'shape'
Did not test it on A1111, as it is a simple controlnet without the need for any preprocessor. Just put the image to inpaint as controlnet input. The denoising strength should be the equivalent of start and end steps percentage in a1111 (from memory, I don't recall exactly the name but it should be from 0 to 1 by default).
On comfyui, just don't use any preprocessor, it has been train to inpaint the white areas, but does not use any real inpaint features of comfyui's nodes.
It was a learning experiment on my side, results can be disappointing, don't put to much expectation or effort on it.
I got this working in Forge by using txt2img (as the author said), and setting the preprocessor to None. I uploaded a 'Mask' which controls where the inpainting will occur.