Midjourney Upgrades Synthetic Image Engine With Consistent Character Generation

Consistent Character Creation in MidJourney
byu/sketner2018 inmidjourney

Generative AI media creator Midjourney has introduced the capability to maintain character consistency across multiple images, a frequent request by users of the service. Though synthetic media engines often face difficulty reproducing the same characters consistently in varied scenes and settings, Midjourney claims to have essentially solved the problem.

Consistent Synthetic Characters

The diffusion models underlying generative AI image creators like Midjourney construct images a pixel at a time from textual prompts and are trained on enormous image databases tagged with relevant text. However, the inherent nature of generative AI in producing unique outputs for each prompt has made continuity almost impossible. Midjourney’s response to this challenge is the introduction of a new tag, “–cref” (character reference), which allows users to specify a character’s features, such as facial appearance, body type, and clothing, by attaching a URL after the tag in their prompts. This tag, available within the Midjourney Discord, aims to replicate the specified character’s attributes closely in new images, facilitating greater narrative coherence in creative projects.

The feature’s effectiveness is enhanced when used with images previously generated by Midjourney. Users can control the degree of similarity between the new and original character images through the “–cw” (character weight) tag, followed by a numerical value ranging from 1 to 100. A higher “cw” value results in greater fidelity to the original character design, while a lower value introduces more variation. Moreover, the platform enables the merging of multiple characters into a single image by using multiple “–cref” tags.

“Hey @everyone @here we’re testing a new “Character Reference” feature today This is similar to the “Style Reference” feature, except instead of matching a reference style it tries to make the character match a “Character Reference” image,” Midjourney founder David Holz wrote on Discorn in announcing the feature. “Drag or paste an image into the imagine bar, it now has three icons. selecting these sets whether it is an image prompt, a style reference, or a character reference. Shift+select an option to use an image for multiple categories. Remember, while MJ V6 is in alpha this and other features may change suddenly, but V6 official beta is coming soon.”

The consistent character feature not only enriches Midjourney’s available toolset for digital artists and creators but also signifies a shift towards more professional AI applications in creative industries. As this feature evolves and improves, it promises to transform Midjourney from a novel ideation aid into an indispensable resource for professional storytelling and visual content creation.

Follow @voicebotaiFollow @erichschwartz

Midjourney’s First Generative AI Art Mobile App is (Almost) All Anime, All the Time

New Midjourney V5 Generative AI Offers Photorealistic Images and Hands With 5 Fingers

Stability AI Upgrades Synthetic Media Engine to Stable Diffusion 3