Control Your Midjourney Creations: A Guide to the New Omni-Reference (V7)

Midjourney has introduced a powerful new feature called Omni-Reference, or simply oref. Think of it as a way to tell Midjourney, "Put this specific thing into my new image." If you're looking to create consistent characters, place specific objects, or repeat scene elements across different generations using the V7 model, oref is what you need.

This guide will walk you through how Omni-Reference works, show you some examples, and give you tips for getting the best results. This is a big step forward for controlling consistency in your AI art.

Getting Started with Omni-Reference

Using Omni-Reference is straightforward, whether you're on the Midjourney website or Discord.

• On the Website: Drag and drop your reference image directly onto the imagine bar where it says "Omni Reference." Make sure your settings are set to V7 first. If you see "Character Reference," you're likely still on V6.

• On Discord: Type `--oref` followed by the URL of your reference image. You’ll need to get the image URL from your uploaded picture.

Your reference image doesn't have to be something you created in Midjourney. You can use external images like photos or artwork. Unlike the older Character Reference in V6, Omni-Reference works very well with images from outside Midjourney, offering much higher consistency, even letting you create images of yourself as a character.

Improved Consistency and Smarts

Omni-Reference is much smarter than its V6 predecessor, Character Reference (cref). It does a better job of understanding elements in your image and placing them correctly. For instance, it handles camera angles much better, avoiding issues where features might appear in strange positions when viewed from different viewpoints. It truly feels like a more accurate system.

Using the Omni Weight Parameter (`--ow`)

You can control how much your generated image should resemble your reference using the `--ow` parameter (Omni Weight).

[LI]Range: The value goes from 0 to 1000.[/LI>

[LI]Default: The default setting is 100. This is a good starting point.[/LI>

[LI]Lower Values (e.g., 0-50): Use lower weights if you want the result to be less similar to your reference. This can be helpful if you want to apply an element from your reference to a new style specified in your prompt.[/LI>

[LI]Higher Values (e.g., 100-1000): Increase the weight for higher similarity. However, very high weights (above 400-500) can sometimes lead to image quality issues or strange results. The best range often seems to be between 50 and 250.[/LI]

[/UL]

Remember that Omni Weight can influence how your prompt's style interacts with the style of your reference image. A high weight might cause the reference image's style to take over, even if you specify a different style in your prompt or using a style reference.

What Can You Reference?

The name "Omni-Reference" suggests its versatility, and it delivers. You can reference many different things:

[LI]Clothing: Use an image of clothing as a reference to put that outfit on different characters.[/LI>

[LI]Scenes/Environments: Reference a scene to bring elements like specific buildings, colors, or environmental features into new images.[/LI>

[/UL]

Handling Multiple Elements & Consistency with Style

A current limitation is that you cannot use multiple Omni-References in one prompt (e.g., one image for a character, another for their shirt). However, there are workarounds:

Placing Two Characters

While direct multiple oref isn't available yet, you can try:

[LI]Using a single reference image that already contains both characters. Midjourney can often recognize them if you describe both in your prompt.[/LI>

[LI]Using the Midjourney editor: Create one character using oref, send the image to the editor, expand the scene, and then add the second character using methods from V6 models like Character Reference (`--cref`) there. You can also use a style reference (`--sref`) of the first character to help match the look when editing.[/LI>

[/UL]

Getting consistent results across many images can be challenging. The Midjourney Automation Suite from TitanXT is designed to help you manage prompts, parameters, and references efficiently, making it easier to achieve uniform styles and elements.

Consistent Style and Storyboards

To build consistent story sequences or apply specific looks, combine Omni-Reference with Style Reference (`--sref`). Using an image as both an oref and an sref can help lock down both the content *and* the look. Remember to adjust the `--ow` value if the style of your oref and sref images clash, and reinforce the desired style in your text prompt.

Advanced Tips for Omni-Reference

Here are a few more things to keep in mind for better results:

• Crop Your Reference: If you only want to reference a character and they are in a busy scene, crop the image tightly around the character. Midjourney considers the whole reference image, and surrounding elements might appear in your results.

• Be Specific in Your Prompt: Always provide context for your reference. If your reference character wears a red and white plaid shirt, mention "a man wearing a red and white plaid shirt" in your prompt, even though the shirt is in the reference. This helps Midjourney understand what elements of the reference you want to include, especially when referencing objects that need to merge with other elements.

• Use Image Prompt and Oref Together: For an even closer match to your reference image, try using it as both a standard image prompt (`--iw`) and an Omni-Reference (`--oref`). This could be useful for getting different camera angles of the exact same setup.

• Parameter Interaction: Be aware that other Midjourney parameters like `--stylize` (`--s`), `--expiration` (`--exp`), mood board references, and style references (`--sref`) can influence how closely your result matches the oref image. If you typically use high stylization, you might need a higher `--ow` or a lower `--s` value.

Managing multiple reference images, parameters, and prompt variations for large projects can become complex. Tools like the Midjourney Automation Suite from TitanXT can streamline this process, helping you experiment faster and maintain consistency.

Conclusion

Midjourney's new Omni-Reference is a significant improvement for creating consistent images. Whether you need stable characters across a series or want to place specific objects with high accuracy, oref in V7 makes it much easier. It's a powerful tool for controlling content and style in your generations.

Experiment with Omni-Reference and see how it can enhance your Midjourney workflow. For advanced control and automation of your Midjourney tasks, consider exploring the Midjourney Automation Suite by TitanXT.