How to Use Midjourney Image Prompts and References

Text prompts can only get you so far. When you need to match a specific visual style, maintain consistency across a series of images, or reference a subject that is hard to describe in words, Midjourney's image-based features fill the gap.

There are three main ways to use images as input in Midjourney: image prompts (pasting a URL directly into your prompt), style references (--sref), and Omni Reference (--oref). Each one works differently and is useful in different situations. This guide covers all three, explains when to reach for each, and includes copyable examples you can try immediately.

Image prompts: using an image URL as input

The simplest form of image reference is pasting an image URL at the beginning of your prompt. Midjourney uses the image as a compositional and visual influence on the output. It does not copy the image. It blends visual information from the image with whatever text description follows.

Image prompts influence composition, color palette, and general structure. They do not reliably preserve specific subjects, faces, or fine details. Think of them as a visual mood board input rather than a copy command.

How to use an image prompt

Upload your reference image to Discord (or any service that gives you a direct image URL), then paste the URL at the start of your prompt before your text description.

https://example.com/reference-photo.jpg a cozy reading nook with warm afternoon light, watercolor illustration --ar 3:2

The URL must be a direct link to an image file. Links to web pages that contain images will not work. You can include multiple image URLs to blend influences from several references.

https://example.com/color-palette.jpg https://example.com/room-layout.jpg minimalist living room, soft daylight, editorial interior photography --ar 16:9

When image prompts are useful

• You have a rough sketch or wireframe and want Midjourney to render a polished version in a specific style.
• You want to borrow the color palette or general composition from an existing photo.
• You are combining two visual ideas and want the model to blend them together.

Limitations of image prompts

• Image prompts do not preserve specific subjects. A photo of your dog will influence the colors and composition but will not generate your exact dog.
• The influence is unpredictable. The model decides how much weight to give the image versus the text.
• You cannot control which aspects of the image the model pays attention to.

Style references with --sref

Style references let you point Midjourney at an image and say 'use this visual style' while ignoring the content of that image. This is the tool you reach for when you want consistent aesthetics across multiple generations.

Unlike image prompts, --sref specifically extracts style information: color grading, texture, rendering technique, and overall mood. It does not try to reproduce the subject matter of the reference image.

Basic --sref usage

a rustic farmhouse kitchen with copper pots and fresh herbs --sref https://example.com/moody-film-photo.jpg --ar 16:9

The reference image in this example is a moody film photograph. The output will be a farmhouse kitchen rendered in that same moody film aesthetic, with similar grain, color grading, and tonal range.

Using --sref for brand consistency

If you are producing a series of images for a brand, campaign, or social feed, --sref keeps the visual language consistent. Generate one image you love, then use it as the style reference for every subsequent image.

a flat lay of artisan bread and olive oil on linen --sref https://example.com/approved-brand-image.jpg --ar 1:1 --stylize 200

This workflow is much more reliable than trying to describe a style in words. The reference image anchors the look, and the text prompt controls the subject matter.

Multiple style references

You can combine multiple --sref URLs to blend styles. This is useful when you want to merge the color palette of one image with the rendering technique of another.

a portrait of a young musician in a recording studio --sref https://example.com/warm-film-tones.jpg https://example.com/gritty-texture.jpg --ar 4:5

Omni Reference with --oref (V7)

Omni Reference is Midjourney's most powerful reference tool. It attempts to preserve specific visual elements from a reference image, including subjects, objects, and characters. Unlike --sref, which only captures style, --oref tries to carry over what the reference image actually shows.

Omni Reference was introduced with V7 and runs on the V7 model. If you include --oref in a V8.2 job, Midjourney will process it using V7 under the hood. This is important to know because the output quality and style characteristics may differ slightly from your usual V8.2 results.

Basic --oref usage

a character standing in a snowy forest clearing, cinematic lighting --oref https://example.com/character-design.jpg --ar 16:9

The model will attempt to place the character from your reference image into the described scene. Results are best when the reference is a clear, well-lit image of the subject against a simple background.

Combining --oref with --sref

You can use --oref for subject consistency and --sref for style consistency in the same prompt. This is useful when you want the same character rendered across different scenes with a consistent visual style.

a character exploring a candlelit cave, dramatic shadows --oref https://example.com/character-ref.jpg --sref https://example.com/painting-style.jpg --ar 16:9

Keep in mind that --oref triggers V7 processing, so the output may look slightly different from pure V8.2 generations even with a style reference applied.

When to use --oref versus --sref versus image prompts

• Use image prompts when you want loose compositional influence and are not trying to preserve anything specific.
• Use --sref when you want to lock in a visual style (color grading, texture, mood) but the subject matter changes between images.
• Use --oref when you need to preserve a specific subject, character, or object across different scenes and angles.

Tips for better --oref results

• Use a clear, high-resolution reference image. Blurry or cluttered references produce inconsistent results.
• Keep the background simple in the reference image. The model has an easier time isolating the subject.
• If the model is not preserving enough detail, try simplifying your text prompt. Long, complex descriptions can compete with the reference image for the model's attention.
• Remember that --oref runs on V7. If you need V8.2 quality for other aspects of the image, consider using --sref for style consistency and handling subject consistency through careful prompting instead.

Combining references with text prompts

The real power of Midjourney's reference system shows up when you combine techniques. A common professional workflow is to use --sref for style consistency across a project while using descriptive text prompts for each individual image.

a wide shot of a coastal village at sunset, fishing boats in the harbor, warm golden light --sref https://example.com/watercolor-style.jpg --ar 16:9 --stylize 250

This gives you a watercolor-style coastal village without having to describe 'watercolor' in words. The reference image does the heavy lifting for style, and the text prompt handles subject and composition.

Common mistakes with image references

• Using --oref when --sref is what you need. If you just want the color palette or rendering style, --sref is simpler and runs on V8.2 natively.
• Expecting exact copies. Midjourney references are influences, not copy commands. The output will always have variation.
• Using low-quality reference images. The model extracts information from what you give it. A pixelated screenshot produces worse references than a clean, high-resolution image.
• Forgetting that --oref uses V7. Plan for slight differences in output quality and style compared to V8.2 generations.

Related guides

• For prompt writing fundamentals, see /blog/writing-better-midjourney-image-prompts.
• For combining references with personalization profiles, see /blog/how-to-combine-references-style-and-personalization-for-design-work.
• For personalization settings that shape Midjourney's default aesthetic, see /blog/how-to-use-midjourney-personalization-profiles.