Flux

Verified against Flux · 2026-08-06

Generate a consistent multi-angle product shot set with Flux.2's multi-reference input

A multi-reference brief for Flux.2 that feeds several existing photos of the same physical product as reference images to generate new angles of that exact product, instead of describing it from scratch and risking a subtly different-looking item in every new shot.

Flux.26 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

REFERENCE IMAGES
front_view.png, three_quarter_left.png, top_down.png — three existing studio shots of the same wireless earbuds case. These 3 images all show the same physical product from different existing angles — treat them collectively as ground truth for its exact shape, proportions, color, and any printed or embossed detail, not as separate style inspirations to blend.

WHAT MUST MATCH EXACTLY ACROSS ALL REFERENCES
the matte charcoal case with a rounded-square shape, the copper LED indicator dot on the front edge, and the hinge seam running along the back. If the references show slightly different lighting or backgrounds from each other, ignore that — extract only the product's physical identity, not its previous lighting setups.

NEW ANGLE TO GENERATE
a side profile view showing the hinge fully open at roughly 45 degrees, camera at eye-level with the case. This angle is not shown in any of the reference images — it needs to be inferred correctly from the product's known geometry across the references you do have, the way a real photographer walking around a physical object would predict what the unseen side looks like from what they can already see.

CONSISTENCY CHECK
Before finalizing, verify the generated angle is physically plausible given the other reference angles — the hinge mechanism and its exact position on the back edge should line up with where it's visible in the three_quarter_left reference — since the biggest failure mode in multi-reference generation is a new angle that looks like a good product photo in isolation but doesn't actually match the same object's proportions from the references.

BACKGROUND AND LIGHTING FOR THIS SHOT
seamless pale grey studio background, soft even top-down light matching the other angles already generated in this set, applied fresh to this angle — consistent with the other new angles being generated in this same set, not necessarily matching whatever background any individual reference image happened to have.

WHAT SHOULD NOT DRIFT FROM THE REFERENCES
Because there is no negative-prompt field, state this positively: the product in the new angle has exactly the same fixed details as every reference image — no resized proportions, no recolored surface, and no altered or invented branding beyond what the references actually show.

BUILDING OUT THE FULL SET
If more angles are needed beyond this one, generate them one at a time using this same reference set rather than trying to request several new angles in a single pass — each new angle should still be checked against the original references individually, and treating this as a repeatable single-angle procedure, run once per needed view, catches inconsistency earlier than a single large batch request would.

CHECK BEFORE FINALIZING
Line this new angle up mentally against the closest existing reference angle and check that the same fixed_product_details would appear where you'd expect if you rotated the physical object that amount — a genuinely consistent new angle should let you predict roughly where a specific feature (a seam, a button, a logo) will land before you even look, and if it lands somewhere that wouldn't make sense given the object's real geometry, treat that as a failed generation rather than an acceptable variation.

OUTPUT
One new image showing the new angle described above of the exact same product as the reference set, styled with the background and lighting described above, ready to sit alongside the other angles in the set as one visually consistent product gallery.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

Flux.2's native multi-reference support — accepting multiple images as simultaneous conditioning input rather than one — is the specific capability this prompt is built around, and it changes the underlying task from "describe a product accurately enough to be recognizable" to "extract a shared physical identity across several already-accurate photos." That is a fundamentally easier and more reliable task, since real proportions and details are being read from pixels rather than reconstructed from a text description that could never fully specify something as fine as an exact hinge seam position. The explicit instruction to extract only physical identity and ignore each reference's individual lighting or background matters because multi-reference conditioning can otherwise blend stylistic elements across the references along with the product identity. If one reference happens to have a warm-toned background and another cool-toned, an under-specified prompt risks generating a muddled average of both instead of a clean, deliberate new styling choice for this specific shot. The consistency-check step — verifying the new angle's geometry against the other references, not just judging it in isolation — targets the actual failure mode of angle-inference tasks: a generated new angle can be a completely plausible-looking product photo that simply doesn't correspond to the same physical object shown in the other angles, with a hinge in a slightly different position or proportions subtly off. This is much harder to catch after the fact than to instruct against up front, since a viewer comparing angles side-by-side notices inconsistency a single image reviewed alone would not reveal. Separating "what must match" — product identity — from "what's generated fresh" — background and lighting for this specific new shot — prevents a different failure. Without that split, the model might either drag one reference's entire background into the new angle, which is inappropriate if the references had inconsistent styling to begin with, or, at the other extreme, treat the whole reference set too loosely and let the product's actual proportions drift along with the styling.

What you get back

A fourth image showing the earbuds case's hinge open at 45 degrees from eye-level, with the hinge position, case shape, and copper LED dot all consistent with the three existing reference angles, styled on the same seamless pale grey background as the rest of the set.

Verified against

Flux Flux.2 · 2026-08-06

Changelog

  • 2026-08-06 Initial publish, verified against Flux.2's multi-reference mode for inferring an unseen product angle from three existing reference photos.

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
Get a clean, ecommerce-ready product shot from Flux.2 with no negative promptsA structured product-photography brief for Flux.2 that produces a genuinely clean, seamless studio background purely through positive description — because Flux has no negative-prompt field to simply tell it "no clutter, no props, no shadows."Flux.2Flux.1 [pro]2026-07-22Brief Flux.2 for an editorial lifestyle photograph, steered by positive descriptionA camera-and-film-aware natural-language brief for Flux.2 that gets a genuine editorial-magazine look — because Flux has no negative-prompt field, unwanted elements have to be steered out through what you describe, not an exclusion list.Flux.22026-07-24Design a scroll-stopping social graphic on Flux.2 with real reserved caption spaceA platform-aware brief for a social campaign graphic on Flux.2 that reserves genuinely clean space for a text overlay added later in a design tool — achieved by describing the empty area explicitly, since Flux has no exclusion field to keep text or clutter out of it.Flux.22026-08-01Keep a character consistent across multiple scenes with Flux.1 KontextA sequential in-context editing workflow for Flux.1 Kontext that locks a character's face, outfit, and proportions across a series of new scenes by editing forward from one reference image each time, instead of re-describing the character from scratch and getting a visibly different person every generation.Flux.1 Kontext [pro]Flux.1 Kontext [dev]2026-07-26
All Flux prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY