Verified against Midjourney · 2026-08-09
Build an Instagram carousel where slide five still feels like slide one
A three-phase workflow for generating a multi-slide Instagram carousel that reads as one continuous aesthetic instead of five separately-styled images fighting each other in the feed, using one anchor slide's style reference across the whole set.
The prompt
Ready to copy — highlighted parts are example details you can swap.
PHASE 1 — ANCHOR SLIDE Generate the carousel's first slide on its own: a hand pouring cold-brew coffee from a glass carafe into a ceramic cup on a marble counter, warm morning light, minimal props, soft neutral color palette, no harsh shadows. This is the slide that sets the color grade, lighting mood, and overall texture every later slide has to match, so pick the result you like best from this run before moving on — do not proceed to phase 2 with a first slide you're only half-happy with, since every later slide is judged against it. PHASE 2 — LOCK THE STYLE Take the image URL of the anchor slide you picked and set it as https://cdn.midjourney.com/jkl012-carousel-anchor.png. This becomes the shared --sref value for every remaining slide in the carousel. PHASE 3 — REMAINING SLIDES For each additional slide, keep warm morning light, minimal props, soft neutral color palette, no harsh shadows and https://cdn.midjourney.com/jkl012-carousel-anchor.png identical and change only the subject: slide 2: coffee beans being scooped into a grinder; slide 3: the finished drink on a breakfast table; slide 4: a person's hands wrapped around the cup outdoors. Generate one slide at a time rather than batching them, and check each new slide against the anchor before generating the next — a small color-grade drift on slide two, left uncorrected, compounds into a visibly mismatched slide four even with the same --sref value applied throughout. STYLE WEIGHT --sw 200 — push this toward the higher end of the range if a slide's subject is visually very different from the anchor slide's subject (a close product shot after a wide lifestyle anchor, for instance), since a big subject-matter gap is exactly when Midjourney's own default aesthetic tends to reassert itself over the reference style. CAROUSEL-SPECIFIC RULE A carousel is judged on completion, not on any single slide in isolation — a viewer who swipes past slide one, two, and three sees them stacked in their peripheral vision before landing on any given slide, so an inconsistency between slides is far more visible here than the same inconsistency would be across five unrelated single posts scrolled past on different days. PARAMETERS --sref https://cdn.midjourney.com/jkl012-carousel-anchor.png --sw 200 --ar 4:5 --v 7 OUTPUT One anchor slide, followed by each additional slide sharing its exact --sref value, together forming a set that reads as one shoot when swiped through in sequence.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A carousel's actual success metric — swipe-through completion — depends on the slides reading as one continuous piece when viewed in rapid sequence, which is a stricter visual bar than five unrelated single posts each judged on their own merits days apart; a viewer's eye catches the edges of adjacent slides in their peripheral vision while swiping, so a color-grade or lighting mismatch between slide two and slide three is far more noticeable here than the identical mismatch would be between two unrelated feed posts. Generating all five slides from five independently-worded prompts, even with identical keyword lists, produces exactly that kind of drift, because each generation still runs through Midjourney's own aesthetic judgment fresh each time with only text to anchor it.
Anchoring every later slide to one chosen first-slide image via --sref, rather than to keywords alone, is what actually closes that gap — --sref conditions each subsequent generation on the anchor's real color grade, lighting character, and texture rather than on the model's independent interpretation of the same adjectives, which is why slide five can depict a completely different subject than slide one and still visually belong to the same shoot. Raising --sw specifically for the slides whose subject matter differs most from the anchor addresses the same style-thinning problem seen across any --sref workflow: the bigger the gap between the anchor's original subject and a new slide's subject, the harder Midjourney's own default aesthetic pulls against the reference style, so an outlier slide often needs a stronger style weight than the rest of the set to stay visually anchored.
Checking each slide against the anchor before generating the next, rather than batch-generating the whole set and reviewing at the end, catches drift while it is still a one-slide problem instead of letting an early inconsistency compound silently across three more generations before anyone notices — by the time a five-slide batch is fully generated, tracing a mismatch back to which slide introduced it is far harder than catching it at slide two.
What you get back
A five-slide carousel where a coffee pour, a bean grind close-up, a finished-drink table shot, an outdoor hands-around-the-cup shot, and a fifth closing shot all share the same warm morning light and neutral palette, reading as one continuous morning-routine story when swiped through.
Verified against
Midjourney v7 · 2026-08-09
Changelog
- 2026-08-09 — Initial publish, verified against Midjourney v7 --sref for multi-slide Instagram carousel consistency.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
