How to turn one fashion image into a 5-asset AI campaign
Build a five-asset fashion campaign from one approved product-reference system: two on-model views, a detail crop, a lifestyle still, and a short vertical Reel.
Lamina Team
Product Team @ Lamina
Start a five-asset AI fashion campaign from one approved product-reference system. From there, build two commerce images, one proof-of-detail image, one editorial still, and one Reel. These are not five prompts spun from a single picture: source photography and verified SKU specifications win every approval call over generated output.
The working unit is bigger than one image. It is a controlled reference package for one SKU: a clean full-garment image plus enough evidence to check color, construction, branding, and the small stuff. That keeps the campaign together without making a generator guess at a collar shape, zipper, print placement, or woven label.
Sequence is where campaigns go sideways. Produce the purchase-critical stills first and approve them at full size; only then make the lifestyle frame and motion derivative. Brandgene’s fashion-production guidance says to plan the image set before generation, while Vantaige names shared model persona, garment reference, and color grade as the controls that keep catalog, ad, social, and video output reading as one campaign.
| Metric | Value | Source |
|---|---|---|
| Planned fashion image-set formats | 6 | brandgene.ioas of 2026-06-25 |
| AI assets generated on Lamina (last 30 days) | 297 | Lamina platform telemetryas of 2026-08-16 |
| Median time to generate an asset | 228s | Lamina platform telemetryas of 2026-08-16 |
| 90th-percentile generation time | 476s | Lamina platform telemetryas of 2026-08-16 |
What belongs in a five-asset fashion campaign?
Give every frame a job: establish silhouette, prove fit, verify one purchase-critical detail, create desire, then add motion. Planned fashion sets commonly use six formats—front PDP, side or back fit, detail crop, lifestyle, paid-social creative, and Story or Reel cover—so five assets make a compact campaign, not a replacement for every PDP view.
Lamina’s last-30-day telemetry shows 297 generated assets. Median generation time is 228 seconds, just under four minutes; the 90th-percentile time is 476 seconds, about eight minutes. That is a generation event, not time to an approved published asset: reference prep, visual review, revisions, export crops, copy, and media setup sit outside those figures. Build the review queue around that difference. Five generated files are not automatically five usable deliverables.
| Asset | Primary job | Required source material | Build rule | Concrete reject example | Source |
|---|---|---|---|---|---|
| 1. On-model PDP hero | Show the complete garment and its silhouette | Clean full front garment image; approved color reference; SKU facts | Use a front or three-quarter full-product pose with an unobstructed neckline, hem, sleeves, and closure. | Reject a hero where hair covers the neckline or the hem is shorter than the approved garment. | photoroom.comas of 2026-06-29 |
| 2. On-model fit view | Show drape, construction, and scale from a second angle | Full side or back reference where available; garment measurements; first approved model image | Reuse the model identity, lighting direction, setting, and grade from the hero. | Reject a fit view that changes stripe spacing, moves a pocket, or conceals a side seam with a bag. | photoroom.comas of 2026-06-29 |
| 3. Macro detail crop | Prove the one feature most likely to affect purchase confidence | Original close-up of the relevant fabric, embroidery, print, closure, hardware, or label | Use the original crop if the source cannot support a truthful generated close-up. | Reject unreadable woven-label text, invented zipper teeth, or a changed embroidery pattern. | masonry.soas of 2026-06-07 |
| 4. Editorial lifestyle still | Place the garment in a campaign world while leaving room for copy | Approved hero or fit view; campaign palette; background and composition direction | Keep the garment prominent and reserve deliberate negative space for a headline or offer. | Reject a lifestyle image where a prop hides the logo or a strong color cast changes the garment color. | vantaige.ioas of 2026-07-22 |
| 5. Vertical Reel | Create a short social derivative from an approved still | Approved on-model or lifestyle still; vertical crop; one approved motion direction | Animate a verified still in 9:16 with restrained movement, then inspect every frame. | Reject a clip with a warping wordmark, shifting buttons, or an arm passing through the garment. | vantaige.ioas of 2026-07-22 |
What belongs in the product-reference package?
At minimum, include a full front image, back image, side image, macro hardware crop, macro label crop, and measured color reference where those views exist. Add the approved product name, SKU, size sample, material notes, and a short invariant list: do not alter the neckline, stripe spacing, print characters, label, hem length, buttons, logo, or color.
Masonry recommends a reference-first test: inspect artwork, pattern, construction, color, material, and proportions at full size before judging the model or pose. That order stops a flattering image from passing just because the person, setting, or light works while the product does not.
Put uncertainty in the brief. Don’t bury it in the prompt. If the source image misses the back closure, supply a back reference; if micro-text is unreadable, use the original detail crop. AI can make textile detail and on-model presentation look believable. It should not manufacture product evidence.
How do you make the five assets repeatable?
Build the SKU truth pack before picking a scene
Put full-garment references, detail crops, verified color, logo artwork, measurements, and invariants in one folder. File names should carry the SKU and colorway. For on-model work, start with a neutral, evenly lit flat lay, cutout, or ghost-mannequin source that shows the entire garment.
Lock the campaign recipe
Set one model identity, pose boundary, setting, lighting feel, color grade, background family, aspect-ratio plan, and output format. Nightjar recommends saving product-shot intent, model, photography style, composition, background, directions, aspect ratio, resolution, and output format as a reusable recipe. Keep the same garment reference in every campaign output.
Generate and approve the two commerce stills first
Make the PDP hero first, followed by the complementary fit view. Check full-size exports against the truth pack before you proceed. Fashion-focused virtual-model workflows can use flat lays or ghost mannequins and offer controls for model, pose, setting, front, three-quarter, and full-body views; use them to show purchase-critical construction clearly.
Build the detail crop and editorial frame from approved evidence
Choose one detail that answers a buying question: rib texture, button, embroidered mark, print, closure, or hardware. Then make the lifestyle still with intentional copy space. The campaign world can shift from collection to collection. SKU truth stays fixed.
Animate an approved still only, then package the outputs
Build a 5–10 second 9:16 Reel from the approved hero or lifestyle frame. Pick one motion cue—a slow push-in, turn, walk, fabric movement, or detail reveal—instead of stacking competing actions. Save the master reference, recipe, prompt or configuration, approval notes, final master, and channel crops together for the next colorway or collection.
How should fashion AI assets be approved before publishing?
Approve in two passes: product truth first, campaign quality second. For pass one, compare the asset at full size with the reference package and inspect silhouette, fit, fabric texture and drape, color, print placement, construction, logo or text, hardware, and proportions. Brandgene flags those details as purchase-critical. A polished image fails if it changes or conceals them.
On the second pass, ask whether the output does its assigned job. A PDP hero must show the whole garment; a detail crop must stay legible; a lifestyle frame needs product prominence and usable copy space; a Reel needs stable geometry through its final frame. Reject misrepresentation. Cosmetic retouching cannot repair an incorrect SKU.
Keep accurate product photography, garment measurements, size information, and clear return policies alongside AI imagery. Perfect Corp recommends AI-generated fashion and try-on imagery as a complement to those customer-facing product facts, never a replacement.
How do you turn an approved still into a Reel?
Animate one approved fashion still with one restrained action. Start with 5–10 seconds in 9:16, choose the first-frame crop before animation, and make the opening readable without waiting for a later reveal. A slow camera push, brief turn, a few walking steps, controlled fabric movement, or a close detail reveal will do the job.
Make several opening variants from the same approved still instead of changing the garment or model for every hook. Open one on the full silhouette, another on sleeve detail, another on the copy-space lifestyle crop. Hold the locked model persona and grade across them so paid social, catalog, and video do not resemble separate shoots.
Review the exported clip frame by frame. Motion creates failures a still hides: logos smear between frames, buttons disappear, color drifts, garment geometry buckles during a turn. Publish only clips that retain the approved still’s product evidence all the way through.
What did Jordache report about AI fashion models?
Jordache executive vice president Shaul Cohen describes AI-model production as a way to reduce fashion-photography cost and time while maintaining the brand’s quality standard. This is Jordache’s own account of using Botika, not a universal benchmark for campaign cost, approval rate, or output fidelity.
Using Botika's AI models has transformed our approach to fashion photography. We've cut costs and reduced our production time significantly, all while maintaining the high-quality standards our brand is known for
| Tier | Price | Included | Best for |
|---|---|---|---|
| Single-SKU proof | 5 generated asset briefs + review | Confirm provider credit use during the difficult-SKU test | Testing one dense-print, logo-heavy, or hardware-heavy garment before a collection rollout |
| Colorway campaign | 5 assets per approved colorway + channel crops | Track generation and revision credits separately | Launching a coordinated PDP, paid-social, and Reel set for one style |
| Collection rollout | 5 assets × approved SKU count + shared recipe maintenance | Use the saved recipe to forecast volume after pilot approval | Teams that need model consistency, marketplace dimensions, and repeatable approvals |
One-SKU pilot
Provider-specific; calculate after the difficult-SKU test2 commerce stills + 1 macro detail + 1 lifestyle still + 1 Reel = 5 asset briefs; add only the provider’s published generation and revision charges after the pilot.
12-SKU collection with one shared campaign recipe
Provider-specific; estimate from the approved pilot’s actual credit use and review hours12 SKUs × 5 asset briefs = 60 asset briefs, plus review of every SKU truth check and final channel crop.
Which workflow fits each asset?
Use a fashion-specific flat-lay or ghost-mannequin-to-model workflow for the two on-model assets, an original crop or product-photo editing workflow for macro proof, controlled compositing for the lifestyle frame, and image-to-video for the Reel. Photoroom describes virtual-model controls for model, pose, setting, custom brand models, and garment-detail preservation. On-Model lists fidelity, model consistency, batch processing, marketplace dimensions, and API support as practical buying criteria for scaled fashion production.
Test one difficult SKU before committing a catalog. Pick dense print, small logo text, unusual trim, reflective hardware, or a complicated closure. You will quickly see whether the tool preserves the details customers use to decide, or merely makes a convincing fashion image.
Can one garment image really support a full campaign?
One clean garment image can launch a campaign. A full reference system is safer. Add front, back, side, macro hardware, macro label, and color evidence whenever the SKU carries details that must remain exact. The five-asset method turns that approved system into a reusable campaign recipe instead of a string of disconnected generations.
Should the detail image be generated?
Generate a detail image only where the source material clearly supports the feature shown. Use the original crop for micro-text, complex embroidery, precision hardware, or any feature generation makes ambiguous. A truthful original detail image beats a beautiful invented one.
How many campaign assets should be approved before making video?
Approve at least the still you will animate before making the Reel. Approving both commerce stills first is stronger: they establish the accepted garment silhouette, color, construction, and model identity. The Reel then stays a controlled derivative rather than becoming another opportunity for the SKU to change.
