General guide for Anon version 0.0.2 idk

  1. Installation
    There's multiple ways of downloading it; I recommend following the this one
    https://rentry.co/UnofficialUnstableGuide
    Follow through step "2 Install Stable Diffusion on Computer"
  2. Finding a favourite checkpoint
    There's hundreds of checkpoints for this purpose I recommed Uber Realistic Porn Merge
    https://civitai.com/models/2661/uber-realistic-porn-merge-urpm
    Scroll down and download the inpainting version,(normal version is for generating, not faking images)
    Put downloaded checkpoint into the Stable-diffusion folder which is in the models folder.
    It's generally structured
    *\webui\stable-diffusion-webui\models\Stable-diffusion
    I have it in my D disk so for me it's
    D:\webui\stable-diffusion-webui\models\Stable-diffusion
  3. Change to correct Stable Diffusion checkpoint
    If you've gotten through https://rentry.co/UnofficialUnstableGuide you should be able to launch the webui1111.
    When launched the top left is a drop down menu, click it and selected the downloaded inpainting model.
    Then click img2img, then inpaint
  4. Learning inpaint
    Basic settings recommend
    Mask blur: 2
    Masked content: latent nothing (If you wanna learn what each do put denoising strength to 0 and generate a picture with each.)
    Inpaint area: Only masked
    Sampling method: DPM++ 2M Karras, Sampling step minimum 15 (DPM++ 2M Karras is the fastest sampler per step, so you can generate till you get a preferred shape, use the same seed then increase steps to get a good result, which would be around 30 steps)
    You can also test: DPM++ SDE Karras, Sampling step 20 (my preferred setting, each sample step is slower, but you "gain" more each step)
    Denoising strength: 0.85
    Rest of the settings don't touch unless you wanna learn (A quick google search and you'll find any info needed)

    Using the inpainting brush
    Just paint the brush over any clothing you want removed

    Prompts
    Basic positive prompts: naked woman, high quality
    Basic negative prompts: blurry, duplicate, deformed, cartoon, animated, render, detached, mutated, twins, sisters, multiple
    My prompt that I use for every nudify I've done:
    Positive: naked woman, (nude photograph), hard nipples
    Negative: shadows, black, dark

  5. Now it's time to generate your first fake image
    Click the big button which says "Generate" and wait however long you need to.
    If you prefer to save images as jpg to save on space, change it in the settings panel.

Questions?
-> Discord AiGenerated#9761

  1. Optional, finding a better checkpoint/model and making it into a inpainting checkpoint/model.
    Once you've gotten a decent hang of it I recommend getting a better model to get less of a porn star look.
    Browse here and find one you like.
    https://civitai.com/
    Personal recommendation: https://civitai.com/models/10961/lazymix-real-amateur-nudes
    Then follow this guide to turn it into a inpainting:
    https://www.reddit.com/r/StableDiffusion/comments/zyi24j/how_to_turn_any_model_into_an_inpainting_model/
  2. Recommend loras, or textual inversions
    Breast torpedo
    https://civitai.com/models/23411/jk-torpedo-breasts-saggy-boobs-experimental
    Cum
    https://civitai.com/models/4376/realcumai-likeness-embedding
    Breastout
    https://civitai.com/models/20355/breasts-out-exposed-breasts-concepthelper
    Vagina
    https://civitai.com/models/7145/uber-realistic-booty-and-vag
  3. Learning about controlnet
    Controlnet is a powerful tool to force where a limb is going to be instead of it being semi random with generating pictures.
    You could create a pose to look exactly like the picture, but then change where the arm is going to be by editing the pose.
    For reference
    https://raw.githubusercontent.com/lllyasviel/ControlNet/main/github_page/p11.png
    https://www.youtube.com/results?search_query=control+net+stable+diffusion
    https://github.com/lllyasviel/ControlNet
Edit
Pub: 19 Mar 2023 19:54 UTC
Edit: 03 Apr 2023 09:40 UTC
Views: 4030