Awesome fakes!!
Bollywood Actress High Quality AI generated celebs only
- Thread starter diffusedbaba
- Start date
Hindi actress fakes and nude edits from Bollywood movies
Actually AI isn't very good at creating faces yet, unless you use the exact LORA trained on that face like this one: Kajal Agarwal - SDXL v1.0 | Stable Diffusion LoRA | Civitai
I am not using any actress specific LORA as of now, but am simply doing a face swap. The regular AI model has some hints about who Kajal Agarwal is and how she looks, etc so they're getting somewhat accurate body type which is all that we need.
Face swap is simple and anybody can do it on any computer. It's extremely simple and don't need any fancy AI for it, but setup might be confusing for many. GitHub - Gourieff/sd-webui-reactor: Fast and Simple Face Swap Extension for StableDiffusion WebUI (A1111 SD WebUI, SD WebUI Forge, SD.Next, Cagliostro)
I'm sure there are easier tools for face swapping. I never bothered as the above extension works so well for me.
Vidya Balan is seriously underrated.
Some attempts with poses using controlnet
These are sometimes so realistic that I feel I'll get into trouble for making these lol
Others (pretty much a full gallery):
Some attempts with poses using controlnet
These are sometimes so realistic that I feel I'll get into trouble for making these lol
Others (pretty much a full gallery):
Looks super real!, although she has aged a bit more than these picturesVidya Balan is seriously underrated.
Some attempts with poses using controlnet
These are sometimes so realistic that I feel I'll get into trouble for making these lol
Others (pretty much a full gallery):
That's by design. Why would I target her 45 year old images when I can just create her from her early 30s whens he was hot.
Built an automated workflow around changing clothes (or removing duh). Way better than the regular img2img workflow.
Uses canny controlnet. If curious about tech, then read here: ControlNet/README.md at main · lllyasviel/ControlNet
Folks in other threads with their terrible img2img inpainting workflow that I see in an old tutorial here and obviously fake undressing should use this instead.
Results: Source and modified
Edit: Noticed that the tshirt text came over. It can be easily removed with some simple blurring in the source image.
Uses canny controlnet. If curious about tech, then read here: ControlNet/README.md at main · lllyasviel/ControlNet
Folks in other threads with their terrible img2img inpainting workflow that I see in an old tutorial here and obviously fake undressing should use this instead.
Results: Source and modified
Edit: Noticed that the tshirt text came over. It can be easily removed with some simple blurring in the source image.
Last edited:
Nice! Can you undress this image of Barkha Singh?Built an automated workflow around changing clothes (or removing duh). Way better than the regular img2img workflow.
Uses canny controlnet. If curious about tech, then read here: ControlNet/README.md at main · lllyasviel/ControlNet
Folks in other threads with their terrible img2img inpainting workflow that I see in an old tutorial here and obviously fake undressing should use this instead.
Results: Source and modified
Edit: Noticed that the tshirt text came over. It can be easily removed with some simple blurring in the source image.
You could probably turn this into an API for noobs like me to use 😁Built an automated workflow around changing clothes (or removing duh). Way better than the regular img2img workflow.
Uses canny controlnet. If curious about tech, then read here: ControlNet/README.md at main · lllyasviel/ControlNet
Folks in other threads with their terrible img2img inpainting workflow that I see in an old tutorial here and obviously fake undressing should use this instead.
Results: Source and modified
Edit: Noticed that the tshirt text came over. It can be easily removed with some simple blurring in the source image.
Can you tweak the body fat% to make her a bit thiccer? The small lower belly fat pouch on women is super hot, and she had it during dirty picture, etc.
Yeah, I am only using something called Control net.
Basically it takes a source image, draws out outlines, depth maps, etc from the image and then I enter prompt to describe the image to combine the two to generate image.
But there are better ways to do things. Something called IP Adapter, which should automatically figure out properties from the image without even needing text, so should be able to pick up properties like skin tone, body type, size, ethnicity without specifying at all.
That's what I am studying about currently.
As for API, that's not needed at all. Most of this is simple stuff and can be done by anyone with maybe an hour or two of setup.
And APIs will never be as reliable as you manually tuning and generating images.
I will write some tutorials here once I have figured it all out. Still learning
Not exactly the same, but similar.
Also slightly lazy workflow, so used just this single image as a source for the face so it's not as sharp as others, but it's close enough.
Details are lost because I can at most work at 1 megapixel while your source image is way sharper.
And boy you have no idea the kind of monstrous workflow I went through for this one, but it's good. Will be useful for other images. I finally have a single image workflow ready. Don't even need to build face models anymore.
Excellent work. Can you make us a tutorial video how you do it. Maybe we can learn and contribute too.
Can you please change her pose and make it look like she's holding her boobsNot exactly the same, but similar.
Also slightly lazy workflow, so used just this single image as a source for the face so it's not as sharp as others, but it's close enough.
Details are lost because I can at most work at 1 megapixel while your source image is way sharper.
And boy you have no idea the kind of monstrous workflow I went through for this one, but it's good. Will be useful for other images. I finally have a single image workflow ready. Don't even need to build face models anymore.
hi... nice thread man.. followed you..
was never a fan of comfyui because its very complex and becomes a spaghetti mess the more components you add to it, but if you master it, the best tool for AI.
was never a fan of comfyui because its very complex and becomes a spaghetti mess the more components you add to it, but if you master it, the best tool for AI.
just a tip, you can stack multiple controlnets on top of each other, canny is good but depth also useful in some casesBuilt an automated workflow around changing clothes (or removing duh). Way better than the regular img2img workflow.
Uses canny controlnet. If curious about tech, then read here: ControlNet/README.md at main · lllyasviel/ControlNet
Folks in other threads with their terrible img2img inpainting workflow that I see in an old tutorial here and obviously fake undressing should use this instead.
Results: Source and modified
Edit: Noticed that the tshirt text came over. It can be easily removed with some simple blurring in the source image.
Yepp.. Just built a workflow for depth. It works better for when clothes are tighter. It's more accurate in those cases.
Super se user pic wonderful work done ✔️Kareena Kapoor, attempt with a floral themed prompt:
pls make a series on Rituparna Sengupta
Guys, please delete your TG account. Someone has leaked edited fakes and case has been filed. All use only this site for communication. Delete TG immediately