如何生成 AI 图像并在不丢失一致性的情况下获得同一角色的不同特写镜头↓

how to generate AI images and get different close up shots of the same character without losing consistency↓

1 - generate your hero image on midjourney. I find that mj v8.2 has been the best model to work with for hyperrealistic portraits 
2 - Use that hero image as a reference with nano banana 2 to get close up shots or new angles of the same shot. In addition to the reference image, prompt it with details of how the model looks. I find that combining a reference image + prompting descriptively helps keep it the most consistent. Here's the prompt for this one: 

~Extreme macro close-up of the same woman from the reference image, keeping her exact identity, pale blue eyes, warm aged skin and silver hair. Tight crop on one eye and eyebrow, straight-on eye-level angle, serene expression. Hyperrealistic detail: pale blue iris with fine striations and a soft catchlight, natural sparse lashes, fine crepey eyelid texture, deep natural wrinkles and age spots, a soft silver brow. Soft even studio beauty light, pale gray background, 100mm macro, razor sharp, photorealistic, not airbrushed, no text.~

The gap between AI generated and real photography is closing a lot faster than most people think. A couple months ago even I thought models still weren't good enough to look human realistic, but now you have such fine control to get things like pores distributed naturally across the skin, individual eyebrow hairs and eyelashes. Natural skin creepiness and peach fuzz along the skin. Take these images and use them as references in videos for strong consistency even in complex shots