One of the holy grails of generative AI storytelling is to create consistency in the characters in images and videos Now, MidJourney has made it a little easier

The new Character Consistency feature works in a similar way to using reference images for style and scene, but focuses on specific character traits found in the source image

One of the main uses for this is to create graphic novels, or to convert MidJourney-generated images into short video clips and animate the lips using something like Pika Labs' Lip Sync, allowing different scenes with the same character It is

As with other MidJourney features, character consistency works by adding a code at the end of the prompt

This works best with MidJourney images as the source, rather than photos or drawings created outside of the AI image generator

You can also set how closely MidJourney will match the source image by following the -cw tag and a number between 1 and 100 at the end of the prompt after -cref

You can also add multiple characters to a scene by specifying the URL of each source image in the second or third --cref Setting -cw very high will capture everything, including clothing, face, and hair Setting it low concentrates only on copying the face so you can change the outfit

I decided to try this with a character named Jack Sloane that I created for a previous AI sound effects project

For the most part, it remained consistent, inheriting his scowl and weathered face between each generation and even keeping his hat and coat attire

For the first few tests, only a few subtle elements were changed, leaving him at his office desk We then lowered the -cw value to 40, changed the prompts and dressed him slightly differently

It removed his jacket, but kept the hat and trademark frown This worked so well that it could be used uncomfortably in a new frame for a graphic novel or short film

Finally, I placed him in the hallway and introduced a second character, Lydia Carver, a woman searching for her missing husband, into the scene with Jack It worked, but not well; two of the four images were of the two Jacks without Lydia, one with the wrong dress and one almost right

Overall, MidJourney did a good job in character consistency It needs some careful prompting and getting the source image correct, but it works

This also makes the era of generated storytelling more accessible and allows for new types of content creation in a variety of mediums [A good writer can visualize his or her creation more easily than ever before, filmmakers can create storyboards to help sell to investors, and others can turn a spur-of-the-moment idea into reality