Consistency in Midjourney: Your Guide to the New Omni Reference Feature

Getting consistent characters and objects in Midjourney images just got easier. Midjourney has introduced a new tool called Omni Reference. This feature replaces the old Cref system. It helps you keep visual elements the same across different images. This works for people, cars, animals, or any other item you need to reference. Here's how it works and what you can expect.

Using Omni Reference

First, make sure you are using Model 7 in your Midjourney settings. You'll also need to decide if you want to use RAW mode or keep personalization off, depending on your needs. To use Omni Reference, you simply drag an image into the Omni Reference box. You can use any image, not just ones made in Midjourney. You can use photos from your computer or uploads.

Setting the Strength

Omni Reference has a strength setting. This setting is also called weight. It controls how much Midjourney tries to match the original image. The value can be anything from 0 to 1000. The default is 100. Midjourney suggests keeping it below 400 in most cases. Starting with the default is a good way to see the initial results.

How Omni Reference Performs with Characters

When testing Omni Reference with people, the results can be quite good. Compared to the older Cref tool, Omni Reference often captures facial features more accurately. Cref sometimes did better with things like hairstyles or clothing in early tests. Increasing the strength setting usually helps make the face and hair more consistent across different images generated from the same reference.

One thing to note is that Omni Reference seems to hold onto the pose and overall scene more tightly than Cref. Since it looks at everything in the reference image, it can make the composition feel less varied. Turning off `style raw` allows Midjourney more creative freedom while still keeping a strong resemblance.

Using detailed prompts helps, especially when you describe the person's appearance or outfit. However, testing showed some issues when using a zoomed-out reference image of a person. It seemed harder for Midjourney to create consistent results without a close-up face in the reference. Using a simpler prompt or adding the person's physical description from the original image creation helped improve consistency in later tests. Close-up, front-facing reference images seem to work best for getting consistent people.

If you want consistent clothing, Midjourney suggests a reference strength around 400. You might also need to add details about the clothes in your prompt for the best match.

Looking to make your Midjourney image generation even smoother? The Midjourney Automation Suite from TitanXT offers tools to streamline your creative workflow. Check it out here: TitanXT Midjourney Automator.

Testing Object Consistency

Omni Reference isn't just for people. It works for objects too. Tests included using an image of a mug created with AI and having a transparent background. Midjourney handled the transparent background image fine as a reference.

For the mug test, the default strength gave okay results, but increasing the strength helped make the mug more consistent. Trying to add text onto the mug was tricky, as Midjourney sometimes struggles with text. Adding a prompt telling Midjourney specifically to put the text on the mug helped. Even with text challenges, the mug design itself showed good consistency at a higher strength setting.

Testing a more complex object, like an alien blaster gun, in a dynamic scene also worked well. Using a strength around 300, Midjourney captured some key details of the complex object impressively, even if it didn't replicate it exactly.

Referencing Outfits and Combos

You can use Omni Reference to put a specific outfit design onto different models. By using the outfit image as the reference, Midjourney did a great job applying the design to different women models, keeping the outfit details consistent.

What about trying to reference a specific person *and* a specific outfit together? Midjourney doesn't currently let you use multiple images as Omni references directly. A test combined a person and an outfit image side-by-side into one composite as the reference. Initially, Midjourney just showed the side-by-side image in the results. To improve this, describing the composite image well using external tools (like GPT or Midjourney's own describe tool) and then using that description in the prompt helped. This approach led to better consistency for both the person and the outfit in the final images generated.

Automate repetitive tasks and get consistent styles with the TitanXT Midjourney Automation Suite. Save time and focus on creating!

Omni Reference with Styles and EXP

When you use Omni Reference alongside a Style Reference or the `stylize` and `exp` parameters, things get interesting. With a default Omni strength, a separate Style Reference might not have much effect. You often need to increase the Style Reference weight to see its influence, which can sometimes reduce the consistency of the object or person being referenced by Omni.

The `stylize` and new `exp` parameters, meant to enhance artistic quality, can also be overpowered by Omni Reference. At default or mid-to-high Omni strength, changing the `stylize` or `exp` values might show almost no difference. To let `stylize` or `exp` have a stronger visual impact, you likely need to lower the Omni weight down to a smaller value, like 25. Only then will you start seeing more of the artistic style from those parameters.

Key Notes and Conclusion

Omni Reference is a powerful tool for achieving consistency in Midjourney Model 7. It works well for both characters and objects. Using close-up reference images and adjusting the strength setting, often keeping it under 400, are good practices.

Keep in mind that Omni Reference is not yet available in the editor. Model 7 fully supports Fast mode, and the quality parameter has been updated for this version.

This new feature makes it significantly easier to generate a series of images featuring the same character or item, which is a big step forward for consistent storytelling or product visualization in Midjourney.

Unlock new possibilities and streamline your workflow with the TitanXT Midjourney Automator. Try it today!