Midjourney revolutionizes image generation – how to use the –oref parameter

The Omni-Reference system (marked as --oref) is a fascinating new feature in Midjourney. This innovative function represents a significant leap forward in controlling exactly what appears in Midjourney-generated images. It’s no longer just a character reference system known from previous versions – it’s a complete tool that allows you to tell Midjourney: “put THIS in my image.”

What is Omni-Reference in Midjourney?

Omni-Reference is an advanced system in Midjourney that allows for precise indication of elements to be included in a generated image. Its versatility is truly impressive, as it works with:

• Characters (human and fictional)
• Objects (from everyday items to advanced technologies)
• Vehicles (cars, spaceships, etc.)
• Non-human creatures (animals, mythical beings, etc.)

Simply put, –oref is a way to tell Midjourney “take this image as a reference and try to recreate its key features in a new context.” Unlike regular style references, which mainly influence aesthetics, Omni-Reference focuses more on preserving specific details and elements of an image.

History of references in Midjourney

Before the introduction of the Omni-Reference system, Midjourney offered more limited possibilities for referencing images. Previous versions mainly allowed the use of references in the context of general style (so-called style reference) or through less precise methods.

The Omni-Reference system represents a significant evolution that primarily:

• Enables more accurate mapping of elements from reference images
• Gives greater control over the degree of reference influence through the omni-weight parameter
• Expands the possibilities of applying references to various objects, not just characters
• Provides better integration with other Midjourney parameters

This new functionality in version V7 significantly raises the level of user control over the image generation process, which is particularly important for professionals needing precisely customized results.

How to use Omni-Reference in Midjourney?

In the Midjourney browser version:

1. Update the model to version V7 in the settings menu
2. Drag an image to the prompt bar and drop it in the area marked “Omni-reference”
3. Use the slider to control the reference strength
4. Write a prompt describing what you want to generate

In Discord with Midjourney:

1. Type your prompt along with the --oref URL parameter where URL leads to an image
2. Use the --ow (omni-weight) parameter to control the reference strength

For example:

Beautiful landscape with mountains, sunset, dramatic lighting --oref mountain_reference.jpg --ow 150

Omni-Weight Parameter (–ow) – key to success in Midjourney

Reference strength is a key element for effectively using the Omni-Reference system in Midjourney. It works in a range from 0 to 1000, where 100 is the default value.

• Low values (e.g., –ow 25): Ideal when you want to change the image style (e.g., from photograph to anime) while preserving only general reference features
• Medium values (default 100): Good balance between fidelity and creativity
• High values (e.g., –ow 400): When you care about precisely rendering a character’s face or clothing details

Important note: Parameters --stylize and --exp compete for influence over the image with Omni-Reference in Midjourney. If you use high stylization or exposure values, you must correspondingly increase the omni-weight value, e.g., character --stylize 1000 --ow 400 --exp 100 --oref person.png.

Interaction with other parameters

It’s worth understanding how –oref interacts with other popular Midjourney parameters:

• –stylize + –oref: Stylization control often requires balancing with omni-weight. High stylization may “overwhelm” the reference if omni-weight isn’t appropriately increased.
• –chaos + –oref: The chaos parameter introduces randomness that can reduce reference influence. Higher omni-weight is recommended with high chaos.
• –quality + –oref: Higher quality values can improve the detail of reference rendering, especially at higher omni-weight values.
• –ar (aspect ratio) + –oref: If the reference image proportions significantly differ from the requested proportions, this can affect rendering. It’s worth adjusting the aspect ratio to match the reference proportions or increasing omni-weight.

My experiences with Omni-Reference in Midjourney

In my practice, I’ve discovered several useful strategies for using this system in Midjourney:

Case 1: Creating consistent characters in different styles

I used a picture of my fictional RPG character as a reference, then generated it in different styles in Midjourney:

Fantasy warrior character with long red hair, armor, anime style --oref warrior.jpg --ow 50

The same warrior character, realistic 3D render, in forest during sunset --oref warrior.jpg --ow 200

With a lower weight (50), I preserved the general character features but allowed for style transformation. With a higher weight (200), I ensured that the face and characteristic clothing elements remained recognizable.

Case 2: Combining objects from different references

I tried to create a composition in Midjourney from two different reference images:

Futuristic car in a metropolis, tall skyscrapers, night scene, cyberpunk --oref car.jpg --oref city.jpg --ow 150

The system handled this surprisingly well, preserving features of both references – the car design and the city atmosphere.

Case 3: Using references to create new concepts

This approach worked great for creating product variants in Midjourney:

Running shoes cyberpunk style, glowing elements, advanced technology, futuristic design --oref sports_shoes.jpg --ow 100 --stylize 400

I received futuristic shoes that maintained the ergonomics of the original but gained a completely new look.

Practical applications of Omni-Reference

The Omni-Reference system finds application in many different creative and professional contexts:

Character and concept development

For artists, game designers, and filmmakers, Omni-Reference offers the possibility of exploring different versions of the same character. You can:

• Test different outfits for an existing character
• Place a known character in different scenarios
• Experiment with different visual styles for the same character

Product design

Designers can use Omni-Reference to:

• Create new versions of existing products
• Explore alternative colors and materials
• Test how a product will look in different contexts

Architecture and interior design

Architects and interior designers can:

• Generate variants of existing buildings or interiors
• Visualize changes in existing spaces
• Experiment with different architectural styles based on existing structures

Adaptations and stylistic transformations

Artists can transform existing artworks:

• From realistic to surrealistic
• From photography to various painting styles
• From modern to historical (or vice versa)

Common challenges and solutions

When using the Omni-Reference function in Midjourney, certain challenges may arise. Here are the most common problems and ways to solve them:

Problem: Too strong attachment to the reference

When the omni-weight value is too high, the model may create images almost identical to the reference, limiting creativity.

Solution:

• Decrease the –ow parameter value
• Add more details in the prompt that will encourage the model to be more creative
• Use the –chaos parameter (e.g., –chaos 25) to introduce additional randomness

Problem: Too weak attachment to the reference

When significant elements of the reference do not appear in the generated image.

Solution:

• Increase the omni-weight value
• Describe specific elements you want to preserve in the prompt
• Reduce the values of –stylize and –exp parameters, which can weaken reference influence

Problem: Distortions or incorrect proportions

The reference is recognizable, but distortions or disproportionate elements occur.

Solution:

• Balance: use medium omni-weight values (100-200)
• Experiment with different –quality values (higher may help)
• Adjust the aspect ratio to match the reference image proportions

Problem: Conflicts between multiple references

When using multiple reference images that compete with each other.

Solution:

• Clearly specify in the prompt which elements from which reference you want to preserve
• Consider using one reference image as the main one (with higher omni-weight)
• Experiment with the order of references – sometimes changing the order can yield better results

Effective prompts for Omni-Reference in Midjourney

Based on my experiments, here are some tips for creating effective prompts in Midjourney:

1. Precisely describe what you want to preserve from the reference: Female character with long blonde hair and red suspenders in anime style --oref woman.jpg --ow 25
2. Always mention elements that should be added: Character holding a sword in a forest, medieval fantasy --oref character.jpg
3. Combine with other Midjourney parameters for better results: Mountain landscape in impressionist style, vivid colors, dramatic lighting --oref mountains.jpg --stylize 600 --ow 200
4. Experiment with different weights: The same prompt with different --ow values can yield surprisingly different results

Advanced prompting techniques

To achieve the best results with Omni-Reference, it’s worth using more advanced prompting techniques:

1. Layered prompting: Portrait of a woman, professional photography, perfect face symmetry, sharp details, clean background --oref portrait.jpg --ow 300 This type of prompt contains both a description of the main object and technical and stylistic details we want to achieve.
2. Negative prompting with the –no parameter: Mythical creature, fantasy art, detailed scales --oref dragon.jpg --ow 200 --no wings, no horns, no fire Allows you to specify what elements you don’t want to see in the generated image, which is especially helpful when working with references containing elements you want to avoid.
3. Color specification: Same character, vibrant color palette, blue and gold outfit --oref character.jpg --ow 150 Focusing on specific color aspects while preserving other reference features.

Compatibility and advanced techniques in Midjourney

Omni-Reference in Midjourney works well with:

• Personalization
• Stylization
• Style references
• Mood boards

Advanced techniques in Midjourney worth trying:

• Combining multiple references: Placing two characters or objects in one reference image (or using two images side by side) and referring to them in the prompt often allows for obtaining both elements in the generated image.
• Controlling composition: Using reference images with a specific composition can help achieve a similar layout in the generated image.
• Mixing styles: Using the –oref parameter for content elements and –style for the overall image aesthetics.
• Iterative approach: Using a generated image as a new reference to gradually get closer to the desired result.

The future of Omni-Reference

The Omni-Reference system is relatively new in Midjourney and can be expected to evolve in future updates. Possible directions for development may include:

• Even more accurate rendering of references
• Smarter interpretation of user intentions
• Better integration with existing parameters
• More options for controlling specific aspects of references
• Functions enabling selective use of only selected elements from references

Summary

The Omni-Reference system in Midjourney is a powerful tool that significantly expands the possibilities of image generation. It allows for precise control over what appears in the image while maintaining creative flexibility.

The key to success is experimenting with different values of the --ow parameter and precisely formulating prompts. The possibilities are virtually unlimited, and the Midjourney system continues to evolve.