Get Consistent Images in Midjourney V7: A Look at Omni-Consistency

Midjourney has introduced a new feature called Omni-Consistency in version 7. This tool helps you create consistent images, whether you want to keep a character, object, or even a pet looking the same across different generations. Forget the old character reference system; Omni-Consistency is here to make things easier. Let's see how it works and what you can achieve.

Getting Started with Midjourney V7 Omni-Consistency

To use this new feature, make sure you have selected version 7 in your Midjourney settings. The setup is simple: upload the image you want to use as a reference. Then, drag and drop this image into the new Omni Reference section on the right side of the interface.

When you add an image, you'll see a slider next to it. This is the Omni Reference strength slider, also called `omniweight`. The default value is 100. This value controls how much Midjourney should follow the reference image compared to your text prompt.

Testing Character Consistency

One of the main goals of Omni-Consistency is to make characters look consistent. Let's test this with a personal photo.

Using a Personal Photo

Using a clear photo of a face produced impressive results right away with the default `omniweight` of 100. The generated images showed a strong likeness to the original photo. This was a significant improvement over past methods, which could be hit or miss.

Changing Camera Angles

The system handled different camera angles well. When asked to generate the same character from a high angle and a side profile, Midjourney largely kept the resemblance high, adapting to the requested view point.

However, a distant view, like a bird's-eye shot where the face is small, struggled a bit. This seems to be a limitation Midjourney sometimes has with small, far-away faces, regardless of the reference.

Referencing a Different Character Photo

Testing with a photo of someone else, specifically one that wasn't a close-up (a woman sitting in a cafe), showed interesting results. With the default `omniweight` (100), the face resemblance varied. However, the press release suggested increasing the weight for clothing consistency. Boosting the `omniweight` to around 400 made the clothing much more consistent with the reference image (skirt, shirt, belt). The face resemblance was decent but not perfect with a distant face in the reference.

Currently, adding multiple images as references for Omni-Consistency isn't possible directly, meaning generating multiple *different* consistent characters in one scene is tricky for now.

Generating consistent characters and scenes can simplify many creative projects. To take your Midjourney work further and automate repetitive tasks, check out the Midjourney Automation Suite by TitanXT. It can streamline your workflow and help you manage complex generation tasks easily.

Exploring Object Consistency

Omni-Consistency also works for making objects consistent.

Complex Objects

Using a reference image of a complex, magical clock-like object produced good resemblance, especially at the default `omniweight` of 100. Increasing the weight to 400 actually led to results less like the original reference object. This suggests that for general object consistency, the default weight might be a good starting point and potentially better than higher values.

Product Photography Test

Testing with a product (a Coca-Cola can) held by a model for a product photo shoot proved challenging. Even with the reference image, Midjourney struggled with rendering the can accurately, especially text and scale, for photorealistic e-commerce style shots. Increasing `omniweight` didn't significantly improve this in initial tests.

Handling Vehicles and Pets

Unusual Vehicles

An unusual vehicle reference (a specific 10-wheel Mustang) also showed limitations. Midjourney got the brand and color right, but consistently struggled to render the unique 10-wheel detail, even with higher `omniweight` values. Adding text in the prompt specifying "10 wheels" alongside the reference helped but still didn't guarantee the result.

If you need unusual details on an object or vehicle to be consistent, you might need to add those details clearly in your text prompt in addition to the reference image.

Non-Human Subjects (Pets)

Testing with a pet photo showed that resemblance improved significantly when the `omniweight` was increased to around 400 or higher. While small details like limbs might still have typical AI oddities (like extra toes), the face consistency for the pet was very good with higher weights, sometimes even copying accessories like a bandana.

Combining Omni-Consistency with Styles

The good news is that Omni-Consistency works with different Midjourney aesthetics, stylization settings, and mood boards.

Using a face reference with Midjourney's default aesthetics and higher stylization kept the face consistent while incorporating a more artistic, dreamy, or colorful look. An experimental parameter (`d-exp`) created results that looked like they belonged in a graphic novel, maintaining face resemblance while adding illustrative qualities.

Combining a face reference with a mood board that had a distinct art style (e.g., ancient Japanese aesthetics) and a style reference also worked well. The face was maintained while being inserted into the mood board's artistic look. Sometimes, especially with faces, the results might lean towards realism by default. To push it more towards the style, you might need to increase the `styleweight` parameter.

Ready to streamline your creative process with Midjourney? Automating repetitive tasks can free up your time for more creative work. Learn more about the Midjourney Automation Suite from TitanXT and see how it can help you manage batches of images, variations, and complex workflows.

Key Takeaways for Using Omni-Consistency

• Face consistency works well, especially with close-up reference photos and photo realism.

• Distant faces in reference images can be less consistent.

• To make clothing consistent with the reference, increase `omniweight` to about 400.

• For general object consistency, the default `omniweight` (100) often works best; increasing it might give worse results.

• Unusual object or vehicle details (like extra wheels) can still be challenging; adding these details to your text prompt can help.

• For pets and non-human subjects, increase `omniweight` to around 400 or higher for better consistency.

• Omni-Consistency works with styles, mood boards, and style references. Boost `styleweight` if you want to prevent defaulting to realism when combining a photo reference with a strong style.

Midjourney V7's Omni-Consistency is a powerful step forward for creating consistent elements in your images. While it has some limitations with very unusual details or distant subjects, it makes keeping characters, objects, and pets consistent significantly easier than before. Experimenting with the `omniweight` is essential to find the best settings for different types of references.

Looking for ways to work more efficiently with Midjourney? The Midjourney Automation Suite by TitanXT offers tools to automate your generations, manage your library, and save time. Explore how automation can enhance your creative workflow today!