Aish as a fairy:
Bollywood Actress High Quality AI generated celebs only
- Thread starter diffusedbaba
- Start date
Hindi actress fakes and nude edits from Bollywood movies
Sonam Kapoor in her royal castle bedroom:
Kangana Ranaut in an arcade:
Mallika Sherawat in a swimming pool:
Similar:
Similar:
Rakul Preet Singh as a butterfly fairy:
Other similar:
And that's all for the day. Let me know if you have any requests. I've figured this workflow out and have optimized it very well
Other similar:
And that's all for the day. Let me know if you have any requests. I've figured this workflow out and have optimized it very well
i want to learn this. can you tell me how?Built an awesome workflow around this type of generation:
Down to 35 seconds/image if I have the face model ready.
Generating face model takes about a minute. Trying to automate that as well
rubina diliak
ComfyUI : GitHub - comfyanonymous/ComfyUI: The most powerful and modular stable diffusion GUI, api and backend with a graph/nodes interface.
A high quality stable diffusion model like RealVisXL: RealVisXL V4.0 - V4.0 Lightning (BakedVAE) | Stable Diffusion Checkpoint | Civitai
Reactor for face swapping: GitHub - Gourieff/comfyui-reactor-node: Fast and Simple Face Swap Extension Node for ComfyUI
It's a fairly simple config once you have it all set up, but getting it to set up might be tricky. Also you need a reasonably good GPU to be able to run these things. Integrated graphics will work but you need at least 16GB RAM in your computer if using integrated graphics and it will be really really slow. I mean really..... slow....
I do not know how to set it up on Windows, so you will have to follow the guides online, there are plenty on YouTube, but sure, experiment around and see what you come up with.
Settings I use for Stable Diffusion model RealVisXL: SDE++ DPM, 4 steps, CFG: 1
Full workflow:
There is also v3.0 Turbo model which is better suited for low VRAM GPUs. The quality will degrade but it's still very usable. Also Turbo model is not good at anything above 512x512 while lightning model can do 1024x1024 at impressive quality.
Last edited:
There you go:
I do need to improve her face model though. I only trained it with like 6 images, as compared to 30+ on everybody else. Fixing it soon.
Other:
Last edited:
There you go. I hope the body type is correct. Don't know how well stable diffusion knows her.
Edit: LMAO her leg dug a hole through the bed in 2nd image. Just AI things.
More of Aish, since the last post was quite silly.
Variant with an apron on:
Variant with an apron on:
Amazing creations! Can you try this- Kriti Sanon nude during a photoshoot (I.e. in a studio with lighting and crew etc. visible, kinda like a BTS)
Beinghuman
Contest Moderator
Staff member
Administrator
Exclusive Faker
Pro Faker
Very Active Member
AI Pro Faker
AI Pro Faker
Moderator of the year 2023
3rd Anniversary
Administrator
Moderator of the year 2022
2nd Anniversary
1st Anniversary
Exclusive Faker
Pro Faker
10K Reactions
1000 Reactions
LV
4
plz use df logoAish as a fairy:
Crew might be tricky with face swaps, so I am trying to avoid it. Here's something for now.
My bad. Didn't know the rules. Just read the sticky thread. I'll add it to my workflow
Amazing stuff, super realistic🔥Crew might be tricky with face swaps, so I am trying to avoid it. Here's something for now.
Can you make semi-nudes too, I.e. in varying states of nudity? For example if we wanted to make her stripping out of a saree, would that be possible?
Okay, I am experimenting with a LORA which has undressing. I'll try to incorporate that in and tag you if I succeed.
Nice one. 2nd image is not a issue.There you go. I hope the body type is correct. Don't know how well stable diffusion knows her.
Edit: LMAO her leg dug a hole through the bed in 2nd image. Just AI things.
An experiment with Munmun Dutta (Babita Ji) swimming underwater:
Other similar:
Other similar: