light-and-ray / sd-webui-lama-cleaner-masked-content

Use lama cleaner before inpainting inside stable-diffusion-webui

Lama cleaner as masked content

This extenstion for AUTOMATIC1111/stable-diffusion-webui adds new value of "Masked content" field in img2img -> inpaint tab. It uses preprocessor from controlnet extension, and allows use it with regular inpainting. So this extension requires Mikubill/sd-webui-controlnet

This option means how to preprocess masked content before pass it into stable diffusion. It useful when you want to remove object in photo. Use inpainting model and denoising straight +-0.4

It also supports my other extension: sd-webui-replacer

Mask:

lama cleaner:

fill:

Others

original:

latent noise:

latent nothing:

Also you can use Lama cleaner in extras tab, if you want to use it without stable diffusion:

For small objects in anime-styled pictures you can also try extension yandere-inpaint-masked-content

Options

Lama cleaner works in 256p resolution, so you can choose upscaler for it.

Go to Settings -> Postprocessing -> Upscaling -> Upscaler for lama cleaner masked content:

Default is ESRGAN_4x. But I recommend to use Waifu2x upscaler from my extension, because it's very fast and good enough for this purpose

You can also add this options in ui, using "Settings for img2img"