Style references are the most useful Midjourney feature for anyone who needs more than one image that looks like it belongs to the same project. Instead of describing a look in words and hoping, you point Midjourney at an image (or a style code) and tell it to borrow the look while you describe new content.
This guide explains what the parameters do, where they trip people up, and a simple workflow for keeping a brand consistent across dozens of images.
What --sref does
--sref tells Midjourney to copy the style of a reference: colour palette, lighting, texture, medium, overall mood. It does not copy the content. If your reference is a watercolour of a lighthouse and your prompt is "a bicycle leaning on a fence", you should get a watercolour bicycle, not a lighthouse.
You can pass three kinds of input:
- An image URL (or an image you upload in the web app and drag into the style reference slot).
- A numeric sref code, such as
--sref 1234567890. Codes are shorthand for a style Midjourney has already learned, shared widely in community libraries. - **
--sref random**, which picks a random style and, after the job runs, replaces "random" with the actual code so you can reuse it.
You can also pass more than one reference and weight them, for example --sref urlA::2 urlB::1 to lean twice as hard on the first.
Style weight: --sw
--sw controls how strongly the reference style is applied. It runs from 0 to 1000, and the default is 100.
- Low values (around 25 to 75) borrow the palette and mood but leave the model room to interpret.
- The default is a sensible starting point for most work.
- High values (300 and up) push hard toward the reference and can start dragging in content from it, like a recurring object or composition.
If your bicycle keeps sprouting lighthouses, lower --sw before you change anything else.
Style reference versions: --sv
Midjourney has rebuilt its style-reference system several times, and --sv selects which version is used. This matters mainly for sref codes: the same code can produce a noticeably different look under a different style-reference version or model version. A code you saved last year may not match its old example images today.
The version numbers and defaults have changed with each model release, and the official docs and community write-ups do not always agree at any given moment, so check the current Style Reference page in Midjourney's docs before relying on a specific --sv value. The practical rule: when you reuse an older code and it looks wrong, try the older --sv value listed in the docs, and always save the full prompt (model version, --sv, --sw) alongside the code, not just the number.
Codes vs image URLs
| Sref codes | Image URLs | |
|---|---|---|
| Where it comes from | Midjourney's style space, found via random or shared libraries | Your own images or ones you have rights to use |
| Shareable | Yes, a short number | Yes, but the image has to stay hosted |
| Predictability | Can shift between versions | Tied to a concrete image you can inspect |
| Brand ownership | Anyone can use the same code | You control the source |
| Best for | Fast exploration, finding a look | Locking a look you already have |
For a brand, images you own are usually the better anchor. Codes are excellent for exploration: run --sref random a few dozen times, keep the codes that fit the brief, then generate your own anchor image and reference that.
Combining with --p and moodboards
Midjourney also has personalization (--p). A personalization profile is built from your ranking of image pairs or from a moodboard of images you upload, and it nudges every image toward your taste. It is broader and fuzzier than a style reference.
They stack. A workable split:
--p(your profile or a brand moodboard) carries general taste: how you like composition, colour saturation, level of detail.--srefcarries the specific look of this project.- The text prompt carries content and composition.
If results get muddy, remove one layer and see which was pulling the wrong way. Check the docs for the exact syntax for selecting a specific profile or moodboard, as it has changed over time.
Keeping a brand consistent: the style line
A reference image alone is not enough to hold a brand together. Small drift adds up: one image is warmer, another has a different lens, a third adds grain. The fix is to pair the reference with a short, fixed style line you paste at the end of every prompt.
A style line has five parts:
- Palette: named colours, not "brand colours". "Deep navy, warm cream, one coral accent."
- Light: "soft diffused daylight from the left, low contrast".
- Composition: "subject on the left third, generous negative space on the right".
- Mood: "calm, confident, unhurried".
- Texture or medium: "matte editorial photography, fine grain".
Then each prompt is only content plus the fixed parts:
a founder reviewing printed charts at a wooden desk, morning,
deep navy, warm cream and one coral accent, soft diffused daylight from
the left, low contrast, subject on the left third with clean negative
space on the right, calm and confident mood, matte editorial photography
with fine grain --sref [your anchor image URL] --sw 100 --ar 16:9Change only the content at the start. When a new image drifts, compare it to the style line word by word to see which element slipped. A prompt that builds this line for your brand in one pass:
A workflow for a full set of brand images
- Explore. Run your brief with
--sref randomand a few hand-picked codes. Aim for breadth, not polish. - Pick directions. Group results into three clearly different directions and show them to whoever approves the work.
- Make an anchor. Generate one strong image in the chosen direction. This becomes your
--srefimage. - Write the style line. Five parts, fixed wording.
- Produce. Hero image, social series, product shots, each as content + style line +
--srefanchor. - Fix pass. Use Vary Region or a re-roll for images that drift, rather than changing the style line.
Steps 1 and 2 are where most time gets wasted. A structured exploration run gets you to three distinct directions quickly:
Run 1 — "editorial minimal to swiss grid": 4 studies, mostly type-led layouts on cream, one with a single coral block. Keep #2. Run 2 — "bold gradient to organic blob": 4 studies, loud and warm; #3 reads premium, #1 and #4 read like a crypto site. Keep #3. Run 3 — "monochrome photo to dual-tone coral + blue": #2 and #4 both strong; #4 has a human figure, which the brief wants. Keep #4. Run 4 — "brutalist to soft 3D": nothing on-brief; the 3D ones drift corporate. Keep none. Run 5 — "magazine collage to clean product UI": #1 is the most distinct thing in the set. Keep #1. Gut cut: 20 → 8. Brief cut (premium + human): 8 → 4. Coverage cut (meaningfully different): 4 → 3. Directions to present: A. Type-led cream with one coral block (Run 1 #2) B. Dual-tone editorial photo with a figure (Run 3 #4) C. Collage with torn-paper edges (Run 5 #1) Move all three into Figma; start high fidelity on B.
- Midjourney
Twenty Direction Studies in Thirty Minutes
Explore the brief's style range in five runs, then curate to three distinct directions for the client.
Style anchor locked: cream #F6F1E7 background, royal blue #1E40AF primary, coral #F26B5B accent, soft diffused window light from the left, rule of thirds, calm, subtle 35mm grain, matte. --sref https://…/anchor.png --sw 200. Result library after one session (8 prompts queued, same anchor): - hero-01 (16:9): founder at standing desk, headline space right — usable - hero-02 (16:9): two colleagues, window behind — light blown out, rerun with "window to the left" - social-01…04 (4:5): four learners in four rooms, same cream and blue — all usable as a set - og-bg-01 (1.91:1): abstract blue and coral blocks on cream — usable - character-01 (1:1): the recurring founder character, front-facing — becomes the --cref portrait Fix pass: hero-01 has a six-fingered hand on the mug. Vary Region on the hand only with "hand holding a ceramic mug, five fingers, natural grip". Upscale Subtle. Ship.
- Midjourney
Midjourney Brand Visual System
Ten prompts from moodboard to style anchor, hero, social series, product shots and a fix pass.
Series work: prints and collections
For a matching set of artworks (a print series, a set of chapter illustrations), the same idea applies with one more lock: a composition rule. "Single central subject, horizon at the lower third, 2:3 portrait" keeps a series feeling designed rather than collected.
[subject for this piece], [fixed style line], single central subject,
horizon at the lower third, generous margin, --ar 2:3
--sref [series anchor] --sw 150Vary only the subject between prints, and lay all of them out side by side before upscaling anything. The odd one out is easy to spot in a grid and hard to spot one at a time.
Common mistakes
- Using a reference image whose content keeps leaking in. Lower
--swor crop the reference. - Saving sref codes without the version info they were made with.
- Describing the style in words and also using a strong
--sref, so the two fight. - Rewriting the style line for each image, which reintroduces drift.
Browse more image prompts and the Midjourney filter. For video in the same visual language, see the Veo 3 prompt guide.