<ul><li>Its works on txt2img, but it sucks, i dont gonna lie xd</li><li>Lora Weight 0.8-1 Stable Diffusion works pretty bad with small things, so I recommend that you upscale the image first and then edit it in the Inpaint tool, in case you can't due to GPU limitations, I highly recommend using the After Detailer extension (<a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">ADetailer</a>). With it you can make images directly in the txt2img, just play with the weight of the Lora (0.75 is the minimum that still makes a decent "byakugan").</li><li>(If your ADetailer doesn't work in img2img, here is a little guide to make it work)</li><li>1- Go to your Stable Diffusion directory, there look for the extensions folder&gt;adetailer&gt;scripts&gt; and there you edit a file named !adetailer</li><li>((E:\STD\ForgeUI\webui\extensions\adetailer\scripts\!<a target="_blank" rel="ugc" href="http://adetailer.py">adetailer.py</a>))</li><li>2- Once inside, using CTRL + F you normally do a quick search, look for a paragraph named (( def is_img2img_inpaint )).</li><li>3- And now just add a # and a return False line and you are done.</li><li>@staticmethoddef is_img2img_inpaint(p) -&gt; bool:# return hasattr(p, "image_mask") and bool(p.image_mask)return False</li><li>4- Save the change and restart your Stable Diffusion.</li></ul>

Files with SHA256: 49147890d83f49d86f8eb60471377471bbb00dc2d9c34add43fb742074996c24