This Inpainting controlnet is outdated.
I highly recommend the Inpainting controlnet at https://huggingface.co/kohya-ss/Anima-LLLite.
This is the Inpainting LLLite ControlNet LoRA for Anima Preview3 Base.
https://github.com/kohya-ss/ComfyUI-Anima-LLLite
https://github.com/Fannovel16/comfyui_controlnet_aux
Two custom nodes are required to use ComfyUI.
The workflow of version 001 and the workflow of version 006 are different.
Description
006
FAQ
Comments (9)
Could the controlnet model be used in forgeNeo (a kind of webui)?
I highly recommend using ComfyUI.
ForgeNeo supports Anima-LLLite, but it is too slow.
it work great for me
good!
Very, very good!
So good, thank you very much!! <3
Hey guys, can anyone share a ComfyUI workflow for the newer kohya-ss/Anima-LLLite · Hugging Face
I dragged the train image in and it already had the workflow embedded in it. The issue is that when I inpaint in the KSampler preview, the inpainting works at first, but during the last 2–3 steps it goes back to the original image I inputted.
I’m testing the inpainting by turning a bald, faceless male into an anime character face like Naruto or other characters.
Work very well thank you very much
Is there any way to make it NOT rescale the image by megapixels, but let it keep its original size? I tried removing the couple nodes related to megapixels, but then the area that is supposed to be fixed (masked) comes out as a mess of shapeless pixels.
I then tried adjusting the flow with Inpaint Crop and Inpaint Stitch nodes, but now the masked part gets very visible seams.
Any help will be appreciated!