Boston, MA – The AI image generating service Midjourney has recently introduced a new feature that allows users to recreate characters consistently across new images, addressing a major challenge faced by AI image generators in the past. Most AI image generators rely on diffusion models, such as Stability AI’s Stable Diffusion open-source image generation algorithm, to piece together images pixel-by-pixel based on text input from users. However, the inconsistency of responses generated by AI applications has been a common issue, with each prompt resulting in something entirely new.
Midjourney’s new feature, known as “–cref” (short for “character reference”), aims to match character features, body type, and clothing based on a URL provided by the user. This feature is particularly valuable for narrative continuity in visual mediums like films, novels, or comic books where consistent characters are essential.
To use the new Midjourney feature, users are advised to start with a previously generated Midjourney image and add the “–cref” tag followed by a URL to match the character in a new scenario. Users can also adjust the level of consistency with the original character by using the “–cw” tag followed by a number from 1 to 100, determining how closely the new image aligns with the reference.
The feature has already garnered attention from artists and creators, who are testing its capabilities. By combining multiple characters into one using multiple “–cref” tags, users can blend information from different images to create unique characters. However, the precision of the technique is limited, as exact details may not be replicated.
As Midjourney continues to refine this feature, it holds the potential to transition from a mere ideation tool to a professional resource for creators. With the ability to generate consistent characters across different scenes and settings, Midjourney is offering a new level of flexibility and precision in AI-generated imagery.