How to Use Midjourney Style References and Save Codes

Use a Midjourney Style Reference to guide the appearance of a new image, or add a numerical style code with --sref. Find codes under Explore > Styles, then keep each tested code with its model version and example outputs. The Style Reference guide↗ (opens in a new tab) explains the image and code workflows.

A list of numbers alone is hard to reuse. fenbs can keep the written style notes, accepted examples, and prompts outside one image session, so another AI can help prepare the next assignment. Your visual review gives those saved entries their value.

How do you find and test a style code?

1. Open Explore > Styles in Midjourney.
2. Use Try Style to test a candidate, Copy to keep its code, or Like to find it later under Likes.
3. Run a plain subject prompt with --sref followed by the selected code. Record the model version you used.
4. Alternatively, try --sref random. After submission, save the numerical code that replaces random.
5. Review your own subjects before marking the code as suitable for an assignment. Save the complete prompt and accepted output with the code.

If you are using a reference image, keep that source file or URL with the same record. Do not expect an upload to become a numerical style code; Midjourney does not provide that conversion. Give the entry a useful human name, such as “soft graphite for editorial sketches,” alongside its exact technical identifier.

What should a style entry actually tell you?

Write what makes the look useful, where it succeeded, and what needs checking. A helpful entry might describe restrained colors, grainy shadows, and broad edges, then point to two reviewed examples. “Looks amazing” gives a future teammate no practical way to decide whether the style fits a new assignment.

Include one concrete use and one boundary. For example: “Approved for small editorial illustrations; review dark backgrounds carefully.” This is your assessment of your own tests. It should not sound like a universal statement about what the code will always produce.

The guide to building a prompt library covers naming and maintaining reusable instructions. Keep the style input separate from those text instructions so you can change a subject without losing the record of its visual treatment.

How do you compare styles without confusing the test?

Pick subjects that resemble your actual work. A children’s book team might test a house, a bicycle, and a person reading. A product team might test an object on a plain background and then the same type of object in use. These are example test choices, not recommended settings from Midjourney.

Hold the subject prompt and model version steady while comparing candidate inputs. Review color, texture, readability, and whether the composition leaves space for the intended layout. Then test the promising style against a different subject to see whether your approval was too narrow.

• Compare outputs at the size you expect people to see, including a small thumbnail if that matters.
• Record both the strongest example and a result that exposes a limitation.
• Separate “interesting” from “approved for this assignment” in your own notes.
• Ask the person responsible for the design to choose the version before recording it as approved.

Use the design critique prompt examples to turn vague preferences into review questions. An assistant can help make the comparison clear, while you decide whether the visual result works for the intended audience.

What changes when you use a different model version?

Style References support versions 6 and later, but codes changed between V6 and V7. The vendor describes --sv 4 or V6 for older looks. Keep the version and complete settings beside every accepted example rather than assuming one number means the same appearance across models.

When you compare another version, add a separate test entry. Give it a date, reuse your subject prompts, and describe the visible differences. Leave the earlier approved entry available until someone chooses the new one. This avoids rewriting the evidence behind an existing design decision.

The Style Reference guide↗ (opens in a new tab) says liking style codes does not train Personalization. Profiles have their own workflow in the Personalization guide↗ (opens in a new tab). If your intended input is a curated collection of images, see how to create and reuse a Midjourney moodboard rather than turning your code list into a different workflow.

How do you avoid a library full of unusable numbers?

Save selectively. A code earns a lasting entry when you have an actual use, a reviewed output, and enough information to repeat the comparison. Keep untested discoveries separate from entries someone has approved. Otherwise the word “saved” starts to imply confidence you have not earned.

Review an entry when the assignment or model version changes. Ask whether the examples are still accessible and whether the written use still describes the work. Remove a recommendation from active use when it no longer fits, while keeping its earlier test history if another asset depends on it.

For a broader view of production checks, use the AI guide for graphic designers. Keep the style choice and remaining checks in fenbs, so the next assistant sees the reason behind the chosen code rather than receiving a mysterious number.

How to keep your style findings with fenbs

1. Create a free account.
2. Give the saved look a clear name that describes the work it suits.
3. Save its code or reference link, model version, complete test prompts, and links to the examples you accepted.
4. Write what to check next and why you chose this look. You can return on any device, and another AI can read the same record before helping you continue.

fenbs keeps the reusable instructions and your findings. It does not turn an uploaded image into a style code or apply that code in Midjourney for you. If Likes already holds everything you need for casual exploration, use it. The separate record helps when visual choices must survive another session or person. Browse the prompts and skills hub for more ways to keep useful instructions.