How to Combine References, Style, and Personalization for Design Work

Midjourney offers three distinct systems for influencing visual output: image references that show what you want, style parameters that control how the image is rendered, and personalization that applies your learned preferences. Each works independently. The real power shows up when you combine them deliberately, using each layer to control a different aspect of the final output.

This guide covers how each system works, where they overlap, where they conflict, and how to stack them for design work that requires precise, repeatable aesthetic control. Every technique includes copyable prompts so you can test combinations directly.

The three systems explained

Image references (image prompts)

Image prompts pass a URL at the start of your prompt. Midjourney uses the reference image to influence subject matter, composition, and color. The reference is compositional: it affects what the image contains and how elements are arranged, not just how they look.

[reference_image_url] a modern office interior, natural light, clean lines --ar 16:9 --no text words

Style Reference (--sref)

Style Reference uses the --sref parameter followed by an image URL. Unlike image prompts, --sref only extracts visual style: color grading, lighting quality, texture treatment, and overall mood. It does not influence subject matter or composition. This separation is what makes it powerful for systems work.

a mountain landscape at sunset, dramatic clouds --sref [style_image_url] --ar 16:9

Personalization (--p)

Personalization applies your accumulated preferences from Midjourney usage and voting. It acts as a global bias toward your preferred aesthetics. Unlike references, it requires no uploaded image and applies broadly across all elements of the generation.

a ceramic bowl on a wooden table, morning light --p --ar 3:2 --no text words

How the systems interact

When you use multiple systems in one prompt, they each influence different aspects of the output. Understanding the hierarchy prevents conflicts where one system overrides another.

• Image prompts influence: subject, composition, spatial arrangement, content
• Style Reference influences: color, lighting, texture, mood, visual treatment
• Personalization influences: subtle preferences across all dimensions, acts as a tiebreaker when other signals are ambiguous
• Text prompt influences: subject matter, scene description, explicit instructions

The general priority order from strongest to weakest: image prompt for composition, --sref for style, text prompt for subject details, and --p as a subtle baseline.

Combination 1: Image prompt plus style reference

This combination gives you compositional control from the image prompt and stylistic control from the style reference. Use it when you have a layout or arrangement you want to replicate but in a completely different visual style.

[composition_reference_url] a workspace with a laptop and coffee, overhead view --sref [dark_moody_style_url] --ar 1:1 --no text words

[composition_reference_url] a workspace with a laptop and coffee, overhead view --sref [bright_airy_style_url] --ar 1:1 --no text words

Same composition reference, different style references. The subject and layout stay consistent while the visual treatment changes completely. This is how you present style options to a client using the same content.

Combination 2: Style reference plus personalization

This pairing anchors the visual style with --sref while letting personalization handle the subtle details that the reference does not specify. It produces results that feel both intentionally styled and personally tuned.

a hand pouring coffee into a ceramic cup, kitchen counter, morning light --sref [warm_editorial_style_url] --p --ar 4:5 --no text words

a stack of books next to a reading lamp, evening atmosphere --sref [warm_editorial_style_url] --p --ar 4:5 --no text words

The style reference controls color grading and lighting. Personalization fills in the gaps: how tightly the frame is cropped, how much bokeh appears in the background, how saturated the colors are within the reference's range. Together they produce more consistent series than either alone.

Combination 3: All three systems together

The full stack uses an image prompt for composition, --sref for style, and --p for personal preferences. This is maximum control and is most useful for design systems where every variable needs to be locked down.

[layout_reference_url] a product on a surface with space for text overlay --sref [brand_style_url] --p --ar 16:9 --no text words letters

The image prompt establishes the layout (product left, text space right). The style reference sets the color grading and lighting. Personalization handles the remaining aesthetic choices. The result is highly predictable and repeatable.

Managing conflicts between layers

Conflicts happen when two systems give contradictory signals. An image prompt showing a bright, white scene combined with a dark, moody style reference will produce inconsistent results because the systems are pulling in opposite directions.

Rules for avoiding conflicts

• Match the general mood of your image prompt and style reference. Do not pair a bright reference with a dark style
• If personalization conflicts with your style reference, remove --p. The explicit reference should win
• Keep your text prompt consistent with your references. Do not describe 'bright sunny day' while referencing a dark moody style
• When results are unpredictable, remove one layer at a time to identify which system is causing the conflict
• If two references fight each other, one of them needs to go. Simpler stacks produce more reliable results

Design workflow: style exploration

When starting a new design project, use these combinations to explore quickly. Generate the same subject with different style references to present options. Then lock the winning reference into your template for all subsequent work.

Exploration phase: testing styles on one subject

a glass water bottle on a stone surface --sref [style_a_url] --ar 1:1 --no text words
a glass water bottle on a stone surface --sref [style_b_url] --ar 1:1 --no text words
a glass water bottle on a stone surface --sref [style_c_url] --ar 1:1 --no text words

Production phase: locked style across subjects

a glass water bottle on a stone surface --sref [chosen_style_url] --p --ar 1:1 --no text words
a ceramic cup on a stone surface --sref [chosen_style_url] --p --ar 1:1 --no text words
a leather notebook on a stone surface --sref [chosen_style_url] --p --ar 1:1 --no text words

Design workflow: client presentations

For client work, the ability to show the same content rendered in different styles is invaluable. Generate three versions of the same scene using different style references and present them as direction options. Once the client picks a direction, that reference becomes your system anchor.

[layout_ref] a modern living room with natural light --sref [minimalist_style_url] --ar 16:9 --no text words
[layout_ref] a modern living room with natural light --sref [warm_cozy_style_url] --ar 16:9 --no text words
[layout_ref] a modern living room with natural light --sref [dramatic_dark_style_url] --ar 16:9 --no text words

Important limitations

Not every Midjourney feature stacks with every other feature. Understanding the boundaries prevents wasted generations.

• Image Prompts, Style Reference (--sref), and Omni Reference are NOT compatible with video generations (--video)
• Personalization (--p) is NOT compatible with video generations
• All style and reference layering must be applied at the image generation stage, before any video conversion
• If you plan to convert images to video later, bake all styling into the source image first
• Very strong style references can override text prompt details. Reduce reference weight if your subject description is being ignored

Building a reference library

Professional design work with Midjourney requires a curated library of reference images. Organize them by purpose: composition references, style references, color palette references, and mood references. Having the right reference ready saves time and prevents the drift that happens when you search for references mid-session.

• Composition references: 5 to 10 images showing layouts you commonly need (hero banners, product cards, social posts)
• Style references: 3 to 5 images per brand or project that define the visual treatment
• Color palette references: images that show your target color relationships clearly
• Mood references: atmospheric images that capture the feeling you want to evoke
• Anti-references: images that show what you do not want, useful for briefing collaborators

Practical checklist for design projects

• Define the target aesthetic before opening Midjourney. What does the final look like?
• Gather 3 to 5 style references that represent the target look
• Test each reference individually to see how Midjourney interprets it
• Build a prompt template with fixed parameters and a subject variable
• Generate a test batch of 5 different subjects with the template
• Evaluate consistency. If drift occurs, strengthen the --sref or simplify the template
• Lock the system and document it for team use
• Save best outputs as future reference material

Next steps

Related reading: /blog/how-to-use-midjourney-image-prompts-and-references covers image prompting in depth. /blog/how-to-use-midjourney-personalization-profiles explains how to build and tune your personalization profile. For turning these consistent images into animated clips, see /blog/how-to-turn-midjourney-images-into-videos (noting that references and personalization must be baked into the source image before video conversion).