Using Your Own Images in Midjourney: Style Reference, Describe, and Image Prompts
Midjourney isn’t only text-in, image-out. You can use your own images to steer style, keep a character consistent, or get prompt ideas. The docs cover this under Using Your Own Images. Here’s a concise overview.
Image Prompts
Image prompts let you add one or more reference images to your prompt. Midjourney uses them to influence content, composition, and color. You can paste an image URL (must be a direct link to the image file, e.g. JPG or PNG) or upload in the web app. Combine with text: e.g. “same style but set in a forest” plus your reference. Details: Image Prompts.
Style Reference (—sref)
Style Reference makes the output match the look and feel of another image—mood, lighting, palette, brushwork. In V7 you use --sref with an image URL and optionally --sw (style weight) and --sv (style version). Great for keeping a consistent “film” across multiple images. See Style Reference.
Omni Reference (—oref) and Character Reference
- Omni Reference (
--oref+--ow) — Puts a person or object from a reference image into your new scene while keeping them recognizable. Useful for the same character in different settings. Omni Reference. - Character Reference — For using the same character across multiple images and scenes. Character Reference.
So: —sref = style/mood; —oref = who or what appears in the image.
Describe: Images Into Prompt Ideas
Describe does the reverse: you give Midjourney an image, and it suggests text prompts that could recreate or remix it. Use /describe in Discord or the Describe tool on the web. Upload your image and pick one of the generated prompts to run with /imagine. Perfect when you have a visual but don’t know how to phrase it. Describe.
Video From Images
Midjourney can also turn images into short videos. See the Video article in the docs for the current video model and how to use it.
Tips for Best Results
- Use direct image URLs (the link should end in .jpg, .png, or similar). Pages with multiple images or text often don’t work.
- For style and character consistency, combine —sref (look and feel) with —oref (person/object) when your plan supports it.
- Use Describe to brainstorm prompts from photos or artwork you like.
For brands and controlled visuals, studios often combine AI tools like Midjourney with traditional CGI, as discussed in posts like CGI vs. Generative AI on the Danthree Studio blog—AI for speed and exploration, CGI for accuracy and rights.