well some pics just never work well.. especially if there are a lot of shadows on the body..
u can try to play with your negative prompts and positive prompts some more or the CFG scale..
Guide Installing Stable Diffusion Webui & Nudifying (Inpainting) Guide
- Thread starter softarchitect
- Start date
-
Welcome to the Fakes / AI / Deepfakes category
When posting AI generated content please include prompts used where possible.
u can try to play with your negative prompts and positive prompts some more or the CFG scale..
Use the inpaint feature from the ControlNet extension, not from the img2img tab, it will completely replace the marked area and not get influence by what is already there. It's much easier to work with and basically works out of the box without any fiddling. If you want to preserve the pose of overpainted areas, ControlNet's OpenPose/Depth/SoftEdge/Canny can help.
For the inpainting with the img2img tab you have to manually paint over the thing you want to change in Photoshop (just rough sketch) and than inpaint that area to get back the level of detail. But given how much easier it is to work with ControlNet, that's hardly worth the effort in most cases.
Last edited:
Most models that aren't directly porn focused will do non-nude just fine. Epicrealism, Photon, Cyberrealistic or Clarity will all give very good results (first one is photorealistic, last one a bit airbrush'y, rest inbetween).
It's missing Python 'insightface' module. If restarting doesn't fix it, you can install it manually:
Open a terminal/cmd in your stable-diffusion-webui folder
If that's doesn't work, you can also try the FaceSwapLab extension, which has the same functionality as ReActor.
Last edited:
Anyone know how to fix the annoying problem generating images which says "A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check."
I generate the images with Stable Diffusion using Docker Desktop as base, so i don't know how to add that "--no-half" commandline. Anybody knows how to fix this?
It's been popping out ever since I updated my stable diffusion, cause first I was using V5 and I had no problem at all generating images of any resolution, but now I'm using V8 and this error keeps showing up every time :(
I generate the images with Stable Diffusion using Docker Desktop as base, so i don't know how to add that "--no-half" commandline. Anybody knows how to fix this?
It's been popping out ever since I updated my stable diffusion, cause first I was using V5 and I had no problem at all generating images of any resolution, but now I'm using V8 and this error keeps showing up every time :(
I can't download the models for fake nudes, could someone give me the correct links?
Could someone give me a prompt, for fake nudes
Could someone give me a prompt, for fake nudes
Last edited:
I get that error whenever I have ControlNet enabled for a while and try to disable it again, solution is to keep ControlNet enabled and just set the weight to zero instead of disabling it. A restart fixes it as well. Changing the prompt can sometimes also help.
is where all the models are, you have to make a login and enable "Mature Content" in the Account Settings if you want to see the porn models. Though even without that you can generate nudes, as many of the normal models will do nudes too, e.g.:
For prompts just check the example images, simple "photo of naked woman" will get you started. For fake nudes you want to look into the FaceSwapLab or ReActor extensions.
Thank you for the guide! For some reason the Realistic Vision InPainting 1.3 keeps failing on download but its working fine with just the uberrealistic for now. I am using this guide to decensor/"Detease" some of my favorites that are censored and/or tease content, I have gotten decent results so far and plan to keep working on them. Anyone know of a way to zoom in when trying to mask? Tried control + mousewheel but that only increases/decreases the size of the brush.
Bought a new GPU (nvidia) which should arrive tomorrow. Since this is my first time using SD/A1111 with Nvidia (ran it locally for a couple of months on AMD), I would like to ask if there are some Tipps you guys could give me.
- Any specific Commandline_args I should use except --xformers for a 4070 12gb?
- Which fork of A1111 is generally the best. The original or is there a better optimized version on github?
If any1 could shoot some knowledge sweet if not just ignore me
- Any specific Commandline_args I should use except --xformers for a 4070 12gb?
- Which fork of A1111 is generally the best. The original or is there a better optimized version on github?
If any1 could shoot some knowledge sweet if not just ignore me
With canvas zoom, pretty sure it's with alt+mousewheel and R to reset.
Anyone recently getting a attributeError with roop unleashed?, with elem_ID not being found, when changing from cuda to cpu in the settings tab?
hey, how can i know what mode\model from Civitai is suitable for inpainting? looking for trans\regular, facial, porn and more
Switch it from "original" to "fill". U also muskd way too much, musk only around the bikini area and not the whole chest for the best results..