Are you tired of AI popping up everywhere you look? Yes? Well, believe me, it's going to get much worse in the near future. With the massive deployment of AI across all sectors, social media, media, and streaming platforms are sure to be flooded with AI content in the future.
So, let's explore what we can do to make generated images look more natural and reduce the annoying AI presence. Normal people in photos don't look like this: Many prompts today look something like this:
"vogue photo of young blonde influencer supermodel woman, studio lighting, neutral background, fashion photo, sony a7r V camera, 85mm f1.4 lens --ar 3:2 --s 50 --v 6.0" But you won't get a realistic photo this way.
The result will be a highly polished and overly perfect photo that looks like it came from AI algorithms. Yes, these types of photos have their uses, but if you want a photo that looks like it's of a real person, try modifying the prompt in the following way:
Add terms like: Average-looking person, Low-quality photo, Selfie, Normal people,...
Don't forget about parameters like --style raw --weird (especially --weird with values around 3-6 will help achieve a more natural look)
Sometimes --stylize (with values below 50) can also be helpful
You can also use another photo (the one you want to emulate) and use its URL and parameter --sref, to guide Midjourney on how the final image should look. Adjust the weight AI takes the reference as important with the parameter --sw (0-1000)
photo of 70 years old man, sitting in the czech pub with glass of beer --s 500 --v 6.0 --ar 3:2
photo of 70 years old man, sitting in the czech pub with glass of beer --s 50 --v 6.0 --weird 3 --style raw
On the left is a generically generated image as imagined by the AI and on the right is an image using the described parameters, which is after all somewhat close to real photographs.