i tried using the "Resize and fill" option but i'm still getting the same size resolution every time.. do i need to change some other kind of settings for it to work?
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.
You'd need to change the height so it's larger than the source (the default "Just resize" would stretch the source image to fit, but "Resize and fill" will keep the original image aspect ratio and fill in the extra space)
Another option is to use the outpainting script, if you scroll to the bottom and select "Outpainting mk2" under script then you can tell it how many pixels to expand, and in which directions:
did you mean cfg scale? i didnt see an sfg but i did set the cfg to 3 and that did make it a lot better. do you have any clue why that happens or why that helped? i do not get the stupid results nearly as often now thanks man
Higher CFG will give your prompts more weight, lower will give stable diffusion a bit more freedom, for inpainting that normally means it picks stuff that matches the source image a bit better
Try the advice Durk Dewdly gave too and make sure you're using an inpainting model. From your screenshot it looks like you're using "realisticVisionV50_v50VAE" but you'll want a model that has an -inpainting suffix for doing inpainting (both realistic vision and URPM have an inpainting version)
oh shit sorry im new to all this i didnt know there was an inpainting specific checkpoint thanks so much haha
Hey can you give us the workflow for these ones? They are really good and not too cartoony. The face looks super realistic in 1, 2 and 6. What was your workflow and what was the prompt for 2 and 6? That would be awesome.
I need to find the original images to get the workflow, but for now I can just say that the main part is I use the ROOP extension for stable diffusion for faces. Honestly I cant go back to normal sort of generated girls or faces now after using it.
You just find a good frontal face pic of a celeb you like and stick it in the extension and it does it so smooth and fast.
And here's some pics so my post isn't just text.
Emma Watson, Lily Collins, Kristen Stewart.
Last edited:
can this ROOP extension work with any face?
or is it just for known celeb's faces that are in its database?
yoo, big fan of your KStew, hoping to see more in the future!
any face, it can be anyone. its a very impressive extension.
it also has its own standalone versions for doing deepfake videos.
All i get is "Something went wrong Connection errored out."
Have i done something wrong? Firefox doesnt even load it
Have i done something wrong? Firefox doesnt even load it
Serious question. Did something change recently with the models? I could of swore I tried these same steps a few months ago, but had pretty terrible results and gave up. Come today, my models are looking far better than what I attempted before
there is literally a emblem on the first post to add emblem + prompt small_tits
how do u use that emblem tho?
I click it and it adds the line "small_tits (1)" to my prompts but it doesn't affect anything...
am i using it wrong? or is it just not working (I've tried to use it several times before on different pics)
the guide dosnt explain what this add-on is or how to use it.. seems lots of others here have this problem too.
can u plz elaborate on this if u know how to make it work?
just put it stable-diffusion-webui\embeddings - then you add to promts small_tits and under the picture, you can see it using the emblem ( )
last line there
before i use the version 1.3 like the first page, all good, then i test using the realistic vision 5.1 vae inpainting,i think its works maybe better? ( im not sure as i only generated small image 500-600pixel ), but it the version 5 seems ignoring the small_tits emebeding, i try with small_tits, breast, small chest, small boobs, A cup, it still generate the big size, and sometimes big areola which i need to get small areola
btw if i already generated the image that i want ( the cup size, the V shape), and dont want to generate new variant one, only wants to add detail or fix the edges blurrynes, how to do this?
btw if i already generated the image that i want ( the cup size, the V shape), and dont want to generate new variant one, only wants to add detail or fix the edges blurrynes, how to do this?
Last edited: