Omni reference (or --oref) helps you get roughly the same character or object across jobs. It allows you to stabilize not just characters, but also… anything? Everything? Think buildings, animals, distinct objects, unique shapes.

Basically, the --oref parameter lets you create a subject based on an image you provide via URL (or attachment, if you use the web). It’s helpful for keeping visual consistency, especially if the prompt includes some stabilizing or anchoring phrases. But, be aware, at time of writing, it can’t be used in the Full Editor (because that’s still v6.1 behind the curtain).

<aside> 💡

The promptcraft challenge with oref: Midjourney is semantically aware of the thing in your attachment and will try to use it archetypally in the new image. For example, if your attached image is boots, Midjourney will struggle if you ask for a boat or a hat in the shape of that boot. It will try to treat the boots as boots. Repurposing objects against their archetype is not easy. To do it, you’ll need to drop --ow to values well below the default of 100 and try some tricks.

</aside>

⭐ Open each section. Keep them all open if you’re using AI with this page.⭐

📦 Example of a pretty standard prompt using --oref .

📦 Distinctive features work best with oref.

📦 Different kinds of prompting do different things with oref.

📦 Semantics, archetypes, and omni reference weight (--ow )

📦 Aspect ratios matter with --oref.

📦 Tutorial: How do I make a character reference sheet?

📦Tutorial: How do I use multiple shapes, objects, or people in one oref?

📦 Other important things to know about --oref.

📦 Troubleshooting --oref