I’ve been working on a story that I would like to put pictures to at some point, and, if possible, I’d like to make consistent character images for it – i.e., pictures of the same character in different settings with different expressions. In a previous post, I mentioned how I used the method of mixing real faces together to create a character that looked a little more consistent over multiple pictures.

Since then, Midjourney has made improvement by leaps and bounds. This is my latest experiment. Essentially, here is what I tried. I created one picture…

Then, I took that seed image by inserting this link that contained the picture and added a descriptor in addition to it. So, for example, I wrote a prompt that included the link and added “woman smiling”:

Then I prompted with the same link but instead put “woman frowning”:

Then the link and “woman angry”:

Then link + woman excited:

Then link + woman shy:

Then I took a few links from all those images and used them as seed links and added the descriptor “standing in a forest”:

And most recently, a few of those images links + photograph + running from robots:

I think there is more work to be done to keep the images more consistent, but I’m happy with this experiment thus far, and I will probably use this method more often.

Thanks again, Midjourney.