The most real and human possible, create an image that looks like a selfie/portrait taken with an iPhone; a casual portrait taken by a friend using the back camera as Subject stands on an outdoor patio near high stools and a grill, wooden vertical posts and a large dark screen behind with a bit of city/greenery peeking through; outfit is a fitted short black button-front romper with a deep neckline, a slim chain necklace with a small blue pendant, and simple bracelets; Subject rests one hand lightly on the cheek while the other arm folds across the waist; bright midday sunlight creates clean shadows and gentle highlights; framing is off-center and a bit crooked with part of the face sharply captured; include micro-details like ribbed fabric stretch, button stitching, faint dust on the screen, chair scuffs, pavement texture, and tiny chrome reflections; specify iPhone wide lens ~24–26mm equiv, ~f/1.6, 1/1200s, ISO 50; portrait 4:5 orientation and aspect ratio; It must not look AI-generated, but rather like it was taken by a real person. There should be no clear subject or specific composition — just a casual, unintentional snapshot. It has slight, natural motion blur. No color grading, no stylization, no cinematic look. The tone must be neutral, not too cool, not too warm; The background must be fully sharp and in focus — no bokeh, no depth-of-field effect, no portrait mode. [DO NOT CHANGE SUBJECT/S' FACIAL FEATURES]

DAMN BEEN STRUGGLING TO MAKE A REAL ONE SO HERE IS THE FINAL OUTPUT IS THERE ANYTHING I CAN CHNAGE ? SUGESSTIONS NEEDED :)

This is for runpod to use RTX 5090 that orginally is really hard to get to work on runpod, but the first 1- Click installer creates a Conda virual environment for the RTX 5080 to work, then it's a wan2.2 T2I Installer and then it's a SDXL NSFW installer. The NSFW workflow have first SDXL and then Flux context auto faceswap and SD 1.5 upscale, and it have a reference image as wildcard or if you want to lock the pose in the reference image you can do that aswell. The workflow doesent need a Lora from you, you just put in a frofile image and it will work flawlessly, just change the prompts for your specific influencer because i ran an goth girl there before i saved it, but i added explaination in the workflows aswell. The hole installation takes around 1h but there is nothing that need to be added. I have a small nipple lora there aswell. Only thing is i couldn't save a hugging face token in github, so it needs to be added manually for the NSFW workflow. But I have one here u can add when you install the SDXL_NSFW installer. Huggingface token: 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 I have been working with this installer for a quite long time because it was really advanced so if you need help with an ordinary installer for a workflow, you can ask me because i have been doing this for quite some time and im starting to be good at it through Vibe coding. Today i made an installer for Qwen images With thoose dash guys poses in a few minutes, so if anyone want an installer for a workflow i can probably fix it 4 u. The hole installation process is explained in my github and its really Boomer friendly. lol Start cooking some images. Enjoy guys 🤗 https://github.com/CryptoAce85/ComfyUI-RTX5090-Runpod BTW, the image is just the SDXL of the workflow i didint take more sceenshots. lol

Guys im keeping the uncensored nano banana mini course up for 24 hours then it will be going away in this group. So nows your final chance if you want it and dont want to purchase the course for $10 extra- the choice is yours It may come back im not really sure, but its been added to the main course. here

really focusing on skin realism and blemishes when it comes to prompting. really starting to see a difference.