How to create consistent characters in Midjourney V6
In this article I will test out 3 different prompts to see if we can create consistent characters in Midjourney. This feature was launched recently for V6, and I’m excited to give it a try. Midjourney has added 2 new tags that we will add to our prompts which is the ‘ — cref’ tag, and the ‘— cw’ tag. The ‘— cref’ tag is to point Midjourney to our image that we want to reference, and the ‘ — cw’ tag is to tell Midjourney how consistent or creative we want the new image to be based on our original reference. There is a scale from 0–100 for Midjourney to be creative or consistent with our reference character. I will place examples below. I hope you learn something from this article, and enjoy the prompts and imagery!
To start, I will use this image below that I made on Microsoft’s Image creator. My goal is to create one prompt with consistency, and another prompt with half consistency and half creativity to see how they vary.
When this image is in Discord, you click on ‘Open Browser’, and then copy the url to reference your image.
The url is what we will reference after our ‘ — cref’ tag.
The first prompt I will try is ‘3d toy of a magician with white gloves, Kathie Olivas, LUKE CHUEH, Disney, Pixar, render, Cinema 4d, keyshot — cref https://s.mj.run/Ut7Xc2AfTNI — cw 100 — r 10'. The ‘— cref’ tag will reference my image, ‘ — cw 100’ will tell Midjourney to be 100% consistent, and ‘ — r 10’ will output 10 sets of images from my prompt.
It did a good job of making the aesthetic and character similar, but I would not consider this the same character as my original. I would expect it to be more accurate to the reference and maybe in a different pose, angle, or background. It seems like a good tool to get variations to the original.
Next I will try the same prompt, but with ’ — cw’ set to 50. This tells Midjourney to be 50% creative, and 50% consistent with the original reference character.
These are more derivative than the ‘ — cw’ tag set to 100. I believe Microsoft Image Creator is better at making these cute images, and Midjourney is better at more realistic images. I wish Microsoft Image Creator had more controls like Midjourney. I will try a more realistic example below to see how Midjourney does.
I am going to use this image below that I got by typing in ‘editorial fashion photograph of stunning african woman as cowgirl, wearing black denim, having sunflower avant-garde hat, overlay the background with film grain effects reminiscent of dana trippe’s retro-inspired photography, in the style of kitsch, hasselblad 500c/m 1993'.
I will now add the ‘— cref’, and ‘— cw’ tag to this prompt to set it to 100 to see if Midjourney outputs a similar woman.
I feel like this example is more spot on to being consistent with the original photo. It is impressive, and I will try one more use cases to see how consistent Midjourney is with styles like comics, or children’s illustrations. This seems like it could be ready to create storyboards for a film.
The next prompt I will use is ‘Create a character illustration for a futuristic, tech-savvy individual with distinct neon accents and wires integrated into their design. The character should be suitable for a cartoon character design in a comic book like spawn mixed with spiderverse style. Consider incorporating unique functional elements such as a black suit that reflects their technological prowess, and ensure that their clothing and accessories showcase a blend of modern and future fashion trends. The setting and surrounding environment should also capture the essence of a future world with innovative structures and advanced technology.’
I was inspired to try this prompt out from a page I found with a lot of great prompts here.
I will take this image and reference it and try the ‘ — cw’ tag set to 100. This time instead of just rewriting the original prompt I will write ‘Have this character in a city background — cref https://s.mj.run/9D4m6tOIrns
— cw 100’
This was the first output that I got, and there are a lot of similarities and differences. This is definitely close enough to storyboard with! This is a fun feature they released, and it’s good enough to create great ideas
and stories.
Thanks for reading this far! Check out my other articles on UX/UI, Figma, art, and other AI tools. Do you have an AI tool I should be checking out? Let me know in the comments. Subscribe for more content soon.
