The Midjourney Omni Reference is a cutting-edge feature introduced in version 7 (--oref), enabling users to incorporate specific characters, objects, or creatures from reference images directly into their prompts. This system expands upon the capabilities of the previous 'character references' feature in version 6, offering greater versatility and precision.
Omni Reference allows users to specify "put THIS in my image," whether it be a character, object, vehicle, or non-human creature. Accessible through both the web interface and Discord, users can upload an image to the Omni Reference bin or use the --oref command with a URL pointing to the desired reference image. Additionally, the omni-weight parameter (--ow) offers control over how strictly the generated image adheres to the reference, with values ranging from 0 to 1000 (default: 100). Lower weights facilitate style changes, such as transitioning from photo realism to anime, while higher weights ensure greater fidelity to specific features like facial details or clothing.
The feature integrates seamlessly with personalization, stylization, style references, and moodboards, allowing users to craft highly tailored outputs. However, parameters like --stylize and --exp can influence the image's adherence to the reference, necessitating adjustments to --ow for optimal results. Users are encouraged to specify key details in their prompts to achieve desired outcomes, such as ensuring an object is held by a character or preserving specific traits during style transfers.
Omni Reference also supports multi-element references, enabling the inclusion of multiple characters or objects from separate images into a single prompt. Though experimental, this functionality opens up new creative possibilities. Midjourney invites users to test the feature, share their results, and provide feedback to refine the system further.
Information shown may be outdated. Found an error? Report it here
