Hi guys, I've been trying and trying to improve to hyper realism and although I haven't reached the stage I wanted, I just wanna test the progress I've made so far. I think I'll stick with Z-image since it's only 6B parameters compared to Flux1 dev, nevermind the new COLLOSAL Flux.2 dev😭.

Prerequisites: First off, you need a body picture and a detailed face picture of your AI influencer (I assume you already have one). Take the example of my Chinese AI influencer below. You'll also need multiple email addresses (you're gonna understand why later). Step 1: Finding a Reference The only thing we need now is an IG post of a famous influencer to replicate. It doesn’t actually need to be famous, and it doesn't have to be from Instagram (Pinterest works too), but it needs to be eye-catching for the boys. I’m gonna use this pic from Bunnie Emmie (third image) to replicate, but I'll be replacing her with my AI influencer. Step 2: Generation Next, go to https://higgsfield.ai/image/nano_banana_2. We’re going to use Nano Banana Pro to make our replica. You can also use Lovart AI (just use something where you have access to Nano Banana Pro lvl 3 censorship—this means sexy images are allowed—but I recommend using Higgsfield). Now, upload the picture you want to replicate first, and then the picture of your AI influencer (veryy important). Then, type this prompt: Keep the lighting, location, and pose from Image 1, but replace the person with the girl from the image where she is wearing [the outfit of your AI influencer] (image 2), ensuring her facial features and body structure are preserved. Also make sure to keep the outfit from Image 1 on the replaced figure. Step 3: Variations When it’s finished, the result might be good or bad. If it’s good, you can make some variations using the Nano Banana Prompt Engineer 5.1 GPT on ChatGPT (https://chatgpt.com/g/g-68ae14451c448191837d0e108a48cf89-nano-banana-prompt-engineer). Upload the picture you just made on Higgsfield to the Nano Banana Prompt Engineer GPT and tell it what you want. Ex prompt to Nano Banana Engineer GPT: Change her pose, make her laying, front view, same influencer type pose, same body proportion, same outfit, no distortion.

Hey guys is there nay tutorial on how to use comfyui? right now i'm using higgsfiled as my workflow and although it's pretty good at generating realism with all the models(nanobananpro, flux pro etc.) i still can't get the consistency, would love to know if there is any type of training on comfyui so i can see if is a better alternative. thank!

Does anyone know any image-to-image generator that allows NSFW now that Wavespeed Seeddream has censorship?

How can I prompt for bytedance/seedream-v4/edit in the best way so that I get the same pose, facial expression, etc.? Unfortunately, it’s not working the way I want.