Style References: Prompt Options for Brand-Consistent Art
Ridgeline & Co. shipped 18 outdoor water bottles across four product lines, and the first batch of illustrations looked like eight different studios had drawn them. The Summit insulated bottle had soft airbrush shading, the Trailhead kids bottle was flat vector, and the Commuter slim bottle came back with a photo-real chrome cap. The art director, Noor, had one approved hero image and a deadline for a catalog and six web banners. The problem was never the subject. It was choosing the right consistency lever in Midjourney and knowing what each one would cost in subject freedom.
This is a fitting problem, not a quality problem. Midjourney gives you four dials that matter here: a style reference image, its weight, a fixed seed, and a Style Tuner code. Below is a table for matching each dial to the job, then a prompt block for every row. Replace the bottle names and keep the parameters. For the parameter syntax itself, the Midjourney style parameter reference is the companion read.
Consistency Lever Selection Table
Noor’s approved hero image is the anchor for the whole series: a matte alpine-teal Summit bottle, desert-clay label band, soft studio light, plain snow-white backdrop. Every row below assumes that anchor exists. If it does not, make it first, because a style reference is only as good as the image you point at.
| Scenario | Recommended option | Why | Cost |
|---|---|---|---|
| First SKU of a new series, one approved hero look | Single style reference at moderate weight | Locks palette and rendering while the subject still changes per bottle | Needs a decent base image; a weak anchor limits everything after it |
| Two approved directions to merge (packaging art plus product photo) | Multiple style references | Blends both looks so the series sits between them | Blends can muddy at high weight; results are less predictable |
| Every SKU must read as one ad template | High style weight plus fixed seed | Near-identical rendering across all 18 bottles | Subjects start to resemble each other; brand detail gets lost |
| Four product lines need variety under one hand | Low style weight | Leaves room for line-specific color and props | Drift, so every batch needs a human curation pass |
| Monthly campaign that must stay on-brand without pasting URLs | Style Tuner code | One short code carries the learned look into every prompt | Tied to the training brief; may not transfer to a brand-new subject |
| Reproducing a previously approved frame closely | Same prompt, same reference, same seed | Reproducible output for a specific approved image | Brittle; a single edited word can shift the whole frame |
Matching the Lever to the Job
Pick by how much the subject has to change, not by how strongly you want the brand to show. When the subject is nearly identical across the set, as it is with 18 bottles, a high weight is fine because there is little subject to protect. When the subject changes a lot across four product lines, a high weight fights you and you get four bottles that look like one bottle in four colors.
The order of operations matters more than the values. Approve one anchor image, generate one second SKU at moderate weight, compare them side by side, and only then raise or lower the weight. Changing three dials at once is how a series drifts. The same discipline applies to lighting consistency, which the Midjourney lighting style guide covers in more depth.
Prompt Blocks for Each Option
1. Single anchor reference. The default starting point for the Summit bottle.
studio product illustration of the Ridgeline Summit 24oz insulated bottle, matte alpine-teal body, desert-clay label band, soft studio light, snow-white seamless backdrop --sref BASE_IMAGE_URL --sw 100 --ar 4:5 --style raw2. Blend of two approved directions. Use when packaging art and photography both have to survive.
studio product illustration of the Ridgeline Commuter slim bottle, slim profile, alpine-teal and desert-clay palette --sref PACKAGING_ART_URL PRODUCT_PHOTO_URL --sw 80 --ar 4:5 --style raw3. High-weight ad template. For banners where every bottle must read as one campaign.
studio product illustration of the Ridgeline Basecamp growler, centered, alpine-teal body, clay band, soft studio light --sref BASE_IMAGE_URL --sw 600 --seed 441902 --ar 4:5 --style raw4. Low-weight line variety. For the kids line, where color and props should shift.
playful product illustration of the Ridgeline Trailhead kids bottle, rounded cap, cheerful accent colors, small trail props, same studio light as the series --sref BASE_IMAGE_URL --sw 40 --ar 4:5 --style raw5. Style Tuner code. For a monthly banner set that should not paste URLs.
studio product illustration of a Ridgeline insulated bottle on a snow-white backdrop, matte finish, soft studio light --style TUNER_CODE --ar 16:9 --style raw6. Fixed-seed reproduction. To regenerate one approved frame for a larger size.
studio product illustration of the Ridgeline Summit 24oz insulated bottle, matte alpine-teal body, desert-clay label band, soft studio light --sref BASE_IMAGE_URL --sw 100 --seed 441902 --ar 4:5 --style rawWhat Each Lever Costs You
Style weight is a trade of control against freedom. Push it toward 600 and the brand is unmistakable, but the Trailhead bottle stops looking playful. Drop it to 40 and you get variety, but no two generated images sit in the same light, so you spend the saved time in curation. A seed buys reproducibility and nothing else: it will not fix a weak reference image, and it will not keep a series consistent on its own.
A Style Tuner code is the fastest option once it exists, but training it locks you to the brief you used. If Ridgeline’s fall catalog adds a lifestyle scene with a hiker, a code trained only on studio bottles will fight the scene. Multiple references blend well but are the least predictable row in the table, so use them for a deliberate merge, not as a shortcut when you cannot decide.
One more cost is invisible until the second season. A series locked at a very high weight is hard to evolve, because the next art director inherits a look with no stated rules behind it. Write down the anchor URL, the light phrase, and the weight next to the catalog, so the run can be repeated in six months by someone who was not in the room when it was set.
A Workable Series Workflow for 18 Bottles
Generate the Summit anchor first and get it signed off. Run the Commuter bottle against that anchor at weight 100, then compare tone and shadow. Once two bottles match, freeze the reference and the light phrase, and generate the remaining SKUs in batches by product line, adjusting only color accents. Keep the approved seed noted next to each line so a rejected frame can be regenerated for a different banner size without re-deciding anything.
Judge every candidate at the size it will actually ship. A 16:9 web banner hides the shading drift that ruins a full-page catalog spread, and a catalog spread hides the badge clutter that ruins a small banner. For the broader prompt structure behind these scenes, the Midjourney prompt guide is where to start.
Keep a single reference sheet of the approved anchor next to the catalog as you build it. When a new SKU arrives mid-season, the fastest path is to paste the anchor URL and the light phrase from the sheet, not to describe the style again from memory. Teams that rely on memory end up with a fifth product line that quietly drifts, and a grid of 18 bottles makes that drift obvious on the page in a way a single image never would.
Sources & Notes
These options were written for an outdoor water bottle brand building an 18-SKU illustration series from one approved hero image, but the selection logic is general: choose the consistency lever by how much the subject must change, then set its weight, then freeze a seed only if you need reproduction. The table is the decision; the prompt blocks are the execution.
Two cautions. Midjourney rendering changes between model versions, so re-check your weights after any version update rather than assuming the same numbers hold. And no weight value replaces art direction: a consistent series still needs one person deciding which frames are on-brand, because the model optimizes for the prompt you wrote, not for the catalog you are assembling.