I’m looking for advice on how to generate realistic photos of people with AI, especially when it comes to making sure the hands, facial features, and overall appearance look lifelike. Additionally, how can I prompt for scenarios where multiple people are in the image, ensuring they don’t look identical or wear the same clothing? Any tips on creating natural variations in posture, expression, and attire would be very helpful.
I know we’re on the OpenAI website, but to be absolutely honest… …your best bet would be to not use ChatGPT/Dalle. Sora is pretty good, but for absolute control I’d personally suggest looking at other tools at this time.
As far as I’m aware, civitai seems to be the place to be right now for imgen.
To keep it on topic, a lot of image generation workflows (sd, flux, etc) use openai’s clip for direction.
It’s a very deep rabbit hole you can get into, and for maximum control you’ll be encountering concepts like embeddings and such.
People are afraid AI kills art, but to be absolutely honest, it’s looking like AI’s just an eldritch paintbrush that takes just as much skill and experience to get what you want.
Thanks for the suggestion. I just signed up for Civitai and tried it out, but honestly, the people I created looked like cartoon characters. I experimented with all sorts of settings, but I’m really aiming for a more realistic look. That’s why I was asking about it in the first place.
I totally agree that there are a lot of workflows and concepts like embeddings that can give more control, but I’m still trying to find the right balance to achieve a more lifelike outcome.