Midjourney just rolled out Omni Reference in V7. This new feature lets you point to a reference image and tell Midjourney to put that specific character, object, or creature into your generated image.
For creatives needing practical, usable results with specific elements, this offers a major productivity boost beyond relying purely on text prompt randomness. It makes getting consistent characters or objects into your generations much more reliable.
You use the --oref parameter with an image link and adjust its influence with --ow (omni-weight), using your text prompt to guide the overall scene and any modifications. It works alongside other V7 features.
Looks like a really practical tool for MidJourney users. The clean UI and reference management feature seem helpful. Nice work and congrats on the launch.
Been playing around with this and find it's still a bit rough around the edges. I like where it's going, but I'd say you have a failure rate of more than 50%. Many times the object that you place in it doesn't look like the object, particularly with people and faces (been experimenting with the omni-weight settings, and found that the default often works best.)
Replies
Hi everyone!
Midjourney just rolled out Omni Reference in V7. This new feature lets you point to a reference image and tell Midjourney to put that specific character, object, or creature into your generated image.
For creatives needing practical, usable results with specific elements, this offers a major productivity boost beyond relying purely on text prompt randomness. It makes getting consistent characters or objects into your generations much more reliable.
You use the --oref parameter with an image link and adjust its influence with --ow (omni-weight), using your text prompt to guide the overall scene and any modifications. It works alongside other V7 features.
CastRecap
WalkSmart
Beautiful, excited to try this.
PictoGraphic
chatgpt's new image model is so dominant it's render everything else useless
Close
Been playing around with this and find it's still a bit rough around the edges. I like where it's going, but I'd say you have a failure rate of more than 50%. Many times the object that you place in it doesn't look like the object, particularly with people and faces (been experimenting with the omni-weight settings, and found that the default often works best.)