Creating Facial Expressions on a Consistent Character in Midjourney, along with Image Prompts and Backgrounds

Systematically Changing Facial Expressions, while maintaining a Consistent Character in Midjourney

UPDATE — April 8, 2023: One can “change the clothes” on a consistent character using the following method:

(1) Create a character, (2) Create the Clothes by itself, (3) Use the image prompts from 1 & 2, and pre-fix “wearing [the clothing]” to a combined prompt!

/imagine 18-year-old Swedish prince named Arthur with striking blue eyes and blonde hair, immersed in a medieval world, combining the whimsy of Disney and Pixar with exceptional detail and atmospheric lighting

/imagine blue and white jacket with gold detailing for a young medieval prince, immersed in a medieval world, combining the whimsy of Disney and Pixar with exceptional detail and atmospheric lighting

/imagine <img-1> <img-2> wearing blue and white jacket with gold detailing, 18-year-old Swedish prince named Arthur with striking blue eyes and blonde hair, immersed in a medieval world, combining the whimsy of Disney and Pixar with exceptional detail and atmospheric lighting

UPDATE — Feb 10, 2023: I continue to maintain that seeds can be used to change facial expressions and other image characteristics. Here is one image (out of the set of 4) I generated this morning using the same seed I had generated several weeks ago:

Original Image

I then prepended “playful smile” to the prompt used to generate the image above. I used the same seed as well, but added — stylize 1000 to the prompt. The resulting image is:

Same Seed as above, but with “playful smile”

Image similarity between the two images is 71.98 % using a website https://www.imgonline.com.ua/eng/similarity-percent.php (I could not find what algorithm they use though).

If I generate four images without the seed and prepended “playful smile” to the prompt above and added — stylize 1000, the resulting set of 4 images is:

Obviously, these images are very different from the first image. To quantify these images compared…