If you're looking for settings in auto111, try these and report back:
model (checkpoint): any photorealistic inpainting model from civitai
mask blur: 3
resize mode: just resize
masked content: original/latent nothing
inpaint area: whole picture (try to find the largest dimension your GPU can handle and still generate quickly. I set height around 1000 and use the width slider to match the picture size)
sampling steps: 25
sampling method: DPM++ 3M SDE Exponential / DPM++ 2M SDE Karras
cfg scale: 1.5
denoise str: 1.0
When drawing the mask on the picture in inpaint tab, mask pretty liberally over all clothing parts. When you get a good output, save it and open it in any type of photoshop with the original picture. Paste your output over the original photo and free transform it to the same size as the original.
model (checkpoint): any photorealistic inpainting model from civitai
mask blur: 3
resize mode: just resize
masked content: original/latent nothing
inpaint area: whole picture (try to find the largest dimension your GPU can handle and still generate quickly. I set height around 1000 and use the width slider to match the picture size)
sampling steps: 25
sampling method: DPM++ 3M SDE Exponential / DPM++ 2M SDE Karras
cfg scale: 1.5
denoise str: 1.0
When drawing the mask on the picture in inpaint tab, mask pretty liberally over all clothing parts. When you get a good output, save it and open it in any type of photoshop with the original picture. Paste your output over the original photo and free transform it to the same size as the original.
Last edited: