Midjourney Omni Reference

Product Introduction

1. Midjourney Omni Reference (--oref) is a feature in the V7 model that enables users to integrate specific visual elements from a reference image—such as characters, objects, or creatures—directly into generated outputs. It replaces and expands upon the character reference functionality in V6, offering broader compatibility with diverse visual elements and enhanced control over stylistic adaptation.
2. The core value of Omni Reference lies in its ability to combine precise visual replication with creative flexibility, allowing users to maintain consistency in recurring elements (e.g., character designs) while adapting them to new styles, compositions, or contexts. This bridges the gap between rigid template-based generation and fully open-ended creative exploration.

Main Features

1. Omni Reference supports multi-element integration, enabling users to reference not only characters but also objects, vehicles, and non-human creatures from a single image or multiple images. For example, a user can generate an image of two characters interacting by referencing separate images for each entity.
2. The omni-weight parameter (--ow) provides granular control over how strictly the output adheres to the reference, with a range of 0–1000 (default: 100). Lower weights (e.g., --ow 25) allow stylistic transformations (e.g., converting a photo to anime), while higher weights (e.g., --ow 400) prioritize exact replication of details like facial features or clothing.
3. The feature integrates with Midjourney’s existing tools, including personalization, stylization (--stylize), style references, moodboards, and the experimental --exp parameter. Users can balance omni-weight against competing parameters like --stylize 1000 by proportionally increasing --ow to maintain reference fidelity.

Problems Solved

1. Omni Reference addresses the challenge of maintaining visual consistency across iterative generations, such as preserving a character’s appearance in different scenes or artistic styles. This eliminates the need for manual post-editing or repetitive prompt adjustments.
2. The tool targets digital artists, game developers, and content creators who require precise control over recurring elements in projects like comic series, game assets, or branded visual campaigns.
3. Typical use cases include transferring a character into a new environment (e.g., medieval warrior in a cyberpunk city), adding specific props (e.g., a unique sword design), or blending multiple references (e.g., combining a creature’s anatomy with a vehicle’s structure).

Unique Advantages

1. Unlike V6’s character-specific references, Omni Reference supports a wider range of elements (objects, creatures) and allows simultaneous multi-element integration, reducing the need for complex prompt engineering or external editing tools.
2. The omni-weight system introduces a dynamic balance between reference adherence and stylistic exploration, a capability absent in most AI image generators that rely on binary “strength” sliders without granular parametric control.
3. Midjourney’s integration of --oref with platform-specific features like moodboards and --exp creates a unified workflow, enabling advanced users to layer references with stylistic and compositional parameters without switching tools.

Frequently Asked Questions (FAQ)

1. How do I adjust the style of a referenced element without losing its core details? Lower the omni-weight (e.g., --ow 25–75) to allow stylistic adaptation while retaining key attributes; combine with --stylize to amplify the desired aesthetic (e.g., --stylize 500 --ow 100).
2. Can I reference multiple elements from different images in a single prompt? Yes, include multiple --oref URLs or upload images side-by-side in the web UI, then specify each element in the text prompt (e.g., “a knight and a dragon --oref knight.png --oref dragon.png”).
3. Why does my high --stylize value override the reference? The --stylize and --exp parameters compete with --oref for influence; increase omni-weight proportionally (e.g., --stylize 1000 --ow 400) to maintain reference fidelity while applying stylistic effects.