santa hat
deerdeer nosedeer glow
Sign In

How do i make an inpainting models

I always wanted to train an inpainting model cause i always use it All the time, i know you need like images with missing part, but i dont realy know what to use or do realy to train it

2 Answers

From my understanding, you can convert any model to an inpainting model through merging. You need the original sd 1.5 model, and the official 1.5-inpainting model.

To perform the merge you select Add Difference mode, put sd1.5-inpainting in the first slot, the model of your choosing in the 2nd slot, and the original sd1.5 in the 3rd slot. Then set the multiplier to 1.0 and merge. Make sure the model is named like "MyModel-inpainting.safetensors".

Maybe there are other ways to do it, but this is the method i've used. If you're talking about making a LoRa for inpainting, which I've seen a few of, I have no idea..

You can also add the inpainting diffusion nodes needed with the model toolkit in Automatic1111, advanced mode.

Your answer