Midjourney unveils Character Reference feature for AI image creation

Midjourney has unveiled a new capability enabling users to maintain a consistent character across various images. Dubbed "Character Reference," this feature, currently undergoing testing, is exclusively accessible on the Midjourney 6 (MJ6) and Niji 6 models. Similar to the tool's existing Style Reference function, Character Reference preserves the essence of the depicted character rather than focusing on stylistic elements.

The announcement came via Midjourney's official X (formerly Twitter) account, emphasising that the feature is only accessible for testing purposes on the MJ6 and Niji models (specifically tailored for anime and illustrative styles). Testing is facilitated on Discord, the platform endorsed by Midjourney. Interested individuals must have a subscription to the platform to utilise the AI image generation capabilities.

Elucidating on the Character Reference feature, Midjourney outlined its usage through the "cref" command, which requires a URL link to a character image. Characters designed within the Midjourney platform yield optimal results. However, the company cautioned that the feature does not support real-life individuals or photographs, as the AI may distort them.

Additionally, the precision of the feature is currently constrained, unable to replicate minor details such as dimples, freckles, or logos on clothing. To exert more control over character consistency, users can employ the "cw" command for Character Weight, adjusting a value from 0 to 100 to enhance consistency, ranging from facial features alone at 0 to encompassing the face, hair, and attire at 100. Notably, Midjourney's efficacy is heavily reliant on precise prompts, particularly with the advent of MJ6, emphasising the importance of grammatical accuracy and prompt specificity.

Users wield control over various aspects of an image through specific commands and prompt language. To ensure character coherence, utilisation of the "cw" command and avoidance of vague prompts are recommended. This feature lends itself to diverse applications, including illustration narratives, digital comic storyboarding, or video production accompanied by background music and audio.