Wan 2.7 Multi-Reference Prompts for Product Campaigns

One reference image can preserve a product. A multi-reference prompt has a harder job: it must preserve the product from one image, borrow the visual language of another, use a person or hand from a third, and still produce a coherent moving shot. That is a different discipline from the single-frame workflow in our Kling 3.0 image-to-video guide. The 20 Wan 2.7 multi-reference prompts below are written for campaign teams that need a launch film, paid-social cut, product-page loop, or localized variant without rebuilding the brief for every asset.

TL;DR

• Give every reference one job: product identity, brand world, talent, environment, or motion.
• State what must remain exact before describing camera movement or atmosphere.
• Use one dominant motion idea per generation; product motion, camera motion, and background motion should not compete.
• Treat the 20 prompts as production briefs. Replace the bracketed fields, keep the reference roles, and set the delivery aspect ratio in the generation controls.
• Generate a clean master first, then adapt it into channel-specific variants instead of changing the product, setting, and motion at the same time.

1. Why Multi-Reference Product Prompts Fail

Most weak multi-reference prompts are not too short. They are poorly prioritized. A prompt may ask the model to copy the bottle from Reference 1, the lighting from Reference 2, the pose from Reference 3, and the camera angle from Reference 4 without explaining which instruction wins when those images disagree. The result can look polished while quietly changing the cap, label proportions, material finish, or package color. For a campaign, those small substitutions matter more than an elaborate camera move.

The fix is to separate identity from appearance and motion. Identity covers the product silhouette, construction, color blocking, logo placement, and any other feature that makes the item recognizable. Appearance covers the set, palette, lighting, wardrobe, and finish. Motion covers what the product, camera, talent, and environment do over time. Write those layers in that order. A concise hierarchy gives the model a clearer decision path than a paragraph of equally weighted adjectives.

Multi-reference generation also exposes conflicts in the source pack. A glossy packshot and a soft editorial mood board can imply incompatible reflections; a hand photographed with a wide lens can make the product scale disagree with a telephoto hero image. Clean references reduce this friction. Use a sharp, unobstructed product image, a style reference with one dominant lighting idea, and a talent or setting image whose perspective can plausibly share the same scene.

2. A Reusable Wan 2.7 Multi-Reference Framework

A Wan 2.7 multi-reference prompt is a short production brief that assigns a stable role to each uploaded image, locks the non-negotiable product details, and then describes one shot from opening frame to closing frame. As of August 2026, reference inputs and available controls can vary by generation mode, so check the current model panel before building a large batch. If you have fewer input slots than planned, combine the style and environment into one clean reference board rather than removing the product anchor.

2.1 Assign Reference Roles Before Writing the Shot

Start with a reference map outside the prose prompt. Reference 1 should usually be the product master because the first requirement is recognition. Reference 2 can define the brand world through palette, surfaces, lighting, and art direction. Reference 3 can supply talent, a hand pose, garment, or scale cue. An optional final reference can define the location or a motion beat, but it should not introduce a second product design. These R1-R4 names are planning labels, not a promise that every interface exposes matching reference IDs. When the model panel does not number inputs, upload them in the same order and identify each one by a visible trait, such as "the teal bottle" or "the warm vanity scene."

Use this compact structure for each generation:

Reference roles: R1 = exact product identity; R2 = brand lighting and palette;
R3 = talent, hand pose, or scale; R4 = environment or motion inspiration.

Create: [deliverable and channel].
Opening frame: [composition].
Action: [one primary motion beat].
Camera: [one camera move and lens feeling].
Continuity: [product details that must not change].
Finish: [lighting, material response, background behavior].
Constraints: [what must not appear or transform].

This structure is intentionally literal. The reference map tells Wan what to preserve; the shot description tells it what to invent. When the output drifts, you can identify whether the failure came from identity, appearance, motion, or constraints instead of rewriting the entire prompt.

2.2 Lock the Product Before Adding Style

Product-lock language should be observable, not emotional. Write “preserve the same cylindrical bottle, black ribbed cap, cream label proportions, and centered mark” rather than “keep the brand perfect.” If the label must remain readable, use a front-facing or shallow three-quarter angle and restrained movement. Do not demand a spinning pack, splashing liquid, shallow depth of field, and exact small typography in the same shot; each adds a different source of instability.

Constraint language works best when it protects the campaign claim. Ask for no extra products, no duplicate caps, no invented accessories, no changed packaging geometry, and no new on-screen text. Keep the constraint block short enough to remain legible as a priority. If a generation repeatedly introduces a defect, simplify the action before adding more negative instructions.

2.3 Replace Every Variable Before Generation

A template becomes copy-ready only after every bracketed variable has been replaced with an observable instruction. Do not leave [PRODUCT], [ACTION], or [FEATURE] in the submitted prompt. Define the product by its visible construction, define the action with one start and finish, and define framing choices explicitly. Use [MARKET] for a real destination rather than a broad region, and replace [LEFT/RIGHT] with the side where approved platform copy will actually sit.

For example, replace [PRODUCT] with “a matte deep teal serum bottle with a black ribbed pump and ivory label,” [ACTION] with “lift the bottle once and return it,” and [LEFT/RIGHT] with the side reserved for approved copy. This turns vague variables into reviewable decisions before generation. Apply the same substitution pass to each numbered prompt below.

3. 20 Copy-Ready Prompts by Campaign Use Case

Each prompt assumes the references are uploaded in the stated order. Replace bracketed details with your product and campaign information. Aspect ratio is listed as a delivery recommendation, but it should be selected in the interface rather than trusted to prompt wording alone.

3.1 Launch Hero Reveal - 16:9

Use R1 for the product master, R2 for campaign lighting, and R3 for the launch set.

Create a premium launch hero video for [PRODUCT]. Use Reference 1 as the exact product identity and preserve its silhouette, cap, label proportions, colors, and material finish. Use Reference 2 only for the [COLOR] lighting palette and Reference 3 for the architectural set. Open on the product in silhouette, then make one slow forward camera move as a controlled edge light reveals the package. End on a clean centered hero frame with generous negative space. No added products, floating text, altered packaging, or sudden camera acceleration.

The expected result is a restrained homepage or presentation opener with one clear reveal. Keep the product large enough to inspect at the end frame.

3.2 Material Macro - 16:9

Use R1 for product construction, R2 for macro lighting, and R3 for the surface material.

Produce a cinematic macro study of [PRODUCT]. Preserve the exact product geometry and surface details from Reference 1. Borrow the narrow highlight, contrast, and shadow softness from Reference 2, and place the product on the [STONE/METAL/FABRIC] surface from Reference 3. Begin with an extreme close detail of [FEATURE], then slide laterally into a three-quarter pack view. Show realistic reflections and tactile texture without changing the label or finish. No liquid unless present in the references, no extra props, no deformed edges, and no generated text.

This version works when material quality is the message. It is less suitable when small package copy must remain readable throughout the shot.

3.3 Ingredient World - 4:5

Use R1 for the pack, R2 for approved ingredients, and R3 for brand color.

Create an editorial ingredient-world video around [PRODUCT]. Keep the product from Reference 1 unchanged and upright. Use only the ingredients visibly shown in Reference 2, arranged with the color and lighting language of Reference 3. Start with a close foreground ingredient moving gently past camera, reveal the product behind it, then settle into a balanced final composition. Motion should feel physical and controlled, with realistic scale and contact shadows. Do not invent ingredients, health claims, badges, label copy, or additional product variants.

The final frame can support a product page or organic feed post. Verify that every visible ingredient is approved for the campaign before publishing.

3.4 Packaging Lineup - 16:9

Use R1 through R3 as separate products from the same approved range and R4 as the set reference.

Create a coordinated range reveal using the exact three products in References 1, 2, and 3. Preserve each product's size relationship, package shape, color, closure, and label layout. Use Reference 4 only for the clean studio set and lighting direction. Begin on Product 1, then make a smooth lateral camera move that reveals Products 2 and 3 in sequence. End with all three products separated clearly in one stable frame. No duplicate packs, swapped labels, morphing containers, invented sizes, or floating typography.

The prompt gives every SKU a place in one continuous move. If the products merge, increase spacing in the references or generate a simpler static lineup.

3.5 Launch Teaser Without Generated Text - 9:16

Use R1 for the product and R2 for the campaign world.

Create a vertical launch teaser for [PRODUCT] without on-screen text. Preserve the exact product identity from Reference 1. Use Reference 2 for the dark-to-bright color transition, atmospheric texture, and premium lighting. Show three connected visual beats in one continuous shot: an abstract close detail, a partial silhouette, then a complete front three-quarter reveal. Keep motion centered within vertical safe areas and finish with a steady product hold. No countdown numbers, logos outside the package, extra objects, packaging changes, or strobing light.

Because the clip contains no generated headline, the image itself remains usable across markets. Platform copy can live in the post rather than inside the generated frame.

3.6 Creator Desk Demonstration - 9:16

Use R1 for the product, R2 for the approved creator or hand, and R3 for the desk.

Create a natural vertical creator-style product demonstration. Use Reference 1 as the exact product. Match the person or hand identity, skin tone, manicure, and sleeve from Reference 2, and use the desk arrangement from Reference 3. Open with the product resting on the desk, then have the hand pick it up once, rotate it slightly toward camera, and return it to the same place. Use soft window light and subtle handheld camera behavior. Preserve realistic grip, finger count, product scale, and label orientation. No face substitution, extra hands, package morphing, or on-screen captions.

Use only talent imagery you own or have permission to use. A simple pick-up action is easier to keep credible than a complex application sequence.

3.7 Morning Routine Insert - 9:16

Use R1 for the product, R2 for the talent, and R3 for the bathroom or vanity environment.

Create a polished morning-routine insert featuring [PRODUCT]. Preserve the product exactly from Reference 1 and the approved talent appearance from Reference 2. Recreate the bright vanity environment and warm daylight from Reference 3. Begin with the product in sharp focus near the mirror, let the talent enter softly in the background, then rack focus as the hand reaches for the product. Keep the camera position stable and the action unhurried. No altered packaging, mirror distortions, duplicate reflections, extra fingers, application claims, or generated text.

This shot provides lifestyle context without needing a full face-to-camera performance. It also leaves room for campaign copy outside the video asset.

3.8 Fashion Accessory Walk-By - 9:16

Use R1 for the accessory, R2 for talent styling, and R3 for location.

Create a vertical fashion campaign shot for [ACCESSORY]. Preserve the exact shape, hardware, stitching, color, and proportions from Reference 1. Use the outfit and talent styling from Reference 2 and the location from Reference 3. Start on a medium close view as the talent walks past camera at a measured pace; track alongside while the accessory remains visible and stable; end on a brief natural stop. Keep realistic fabric and strap movement. No duplicated hardware, changing logos, fused hands, wardrobe changes, background crowds, or speed ramps.

The prompt prioritizes product visibility over choreography. For small accessories, shorten the walk and keep the framing tighter.

3.9 Fitness Product Action - 4:5

Use R1 for the product, R2 for the athlete, and R3 for the training environment.

Create an energetic but controlled campaign video for [FITNESS PRODUCT]. Keep the exact product construction, color, and markings from Reference 1. Match the approved athlete and wardrobe from Reference 2, and use the training space from Reference 3. Show one clear action: [ACTION]. Use a low three-quarter tracking camera that keeps the product unobstructed through the movement. Preserve believable weight, contact, and body mechanics. No impossible performance claims, duplicated limbs, changing equipment, added branding, or aggressive camera shake.

One repeatable movement is more useful than a montage when the product must remain identifiable. Avoid implying outcomes the product cannot substantiate.

3.10 Food or Beverage Pour - 9:16

Use R1 for the package, R2 for the serving vessel, and R3 for the setting.

Create a vertical serving ritual for [FOOD OR BEVERAGE PRODUCT]. Preserve the exact package from Reference 1 and the approved glass, cup, or bowl from Reference 2. Use the table setting and light from Reference 3. Begin with a clean product-and-vessel composition, then show one realistic pour with natural liquid weight, condensation, and contact shadows. End with the package upright beside the filled vessel. No overflowing liquid, changing package shape, duplicate containers, invented garnish, unsupported claim text, or floating ingredients.

The shot is designed to feel complete without added titles. Confirm that the depicted serving size and garnish match the approved campaign brief.

3.11 Marketplace Feature Turntable - 1:1

Use R1 for the product, R2 for the approved angle, and R3 for the background tone.

Create a clean square marketplace product video. Preserve every visible design detail from Reference 1. Use Reference 2 to define the approved three-quarter viewing angle and Reference 3 for the neutral background color and shadow softness. Keep the product centered while it makes one slow partial turn, revealing [FEATURE] without hiding the front for long. Maintain consistent scale, geometry, and surface finish from first frame to last. No decorative props, extra variants, altered labels, wobble, morphing, borders, or text overlays.

This is a lower-risk template for a product-page motion tile because it removes narrative distractions. If exact label copy is critical, replace the turn with a shallow camera arc.

3.12 Feature Close-Up - 1:1

Use R1 for the full product, R2 for the feature detail, and R3 for lighting.

Create a square feature-detail video for [PRODUCT]. Keep the full product identity from Reference 1 and reproduce the exact [HINGE/TEXTURE/CONTROL/CLOSURE] shown in Reference 2. Use the crisp directional lighting from Reference 3. Open on the complete product, push in slowly to the feature, show one physically plausible interaction, then hold on the detail. Preserve assembly, material thickness, and scale. No exploded parts, invented mechanisms, changing colors, extra fingers, claim text, or technical labels.

The physical interaction should prove only what is visible. Avoid turning an aesthetic detail into a performance demonstration without evidence.

3.13 Unboxing Sequence - 4:5

Use R1 for the closed box, R2 for the internal arrangement, and R3 for the approved hands and surface.

Create a refined unboxing video for [PRODUCT]. Preserve the exact exterior packaging from Reference 1 and the approved interior layout from Reference 2. Match the hands, sleeves, and tabletop from Reference 3. Begin with the sealed box centered, open it in one smooth natural action, reveal the interior without rearranging its contents, and end on a clear overhead three-quarter view. Keep folds, inserts, product count, and scale consistent. No tearing, extra items, missing inserts, warped hands, invented cards, or readable generated copy.

This prompt favors continuity over fast cuts. If the box mechanism is unusual, provide a reference that clearly shows how it opens.

3.14 Bundle Build - 16:9

Use R1 through R3 for bundle components and R4 for the final composition.

Create a wide campaign video that assembles the approved [BUNDLE NAME]. Use References 1, 2, and 3 as exact identity anchors for each component, preserving their individual scale, colors, and packaging. Use Reference 4 for spacing, backdrop, and lighting only. Start with an empty set; let each component enter separately with restrained physical motion; finish on the exact bundle composition. Products must remain distinct and upright. No duplicates, substitutions, collisions, label swaps, floating text, or additional gift items.

The empty-to-complete progression makes the bundle easy to understand. Keep entry motions simple so the final product identities receive most of the visual attention.

3.15 Retail Shelf Focus - 16:9

Use R1 for the hero SKU, R2 for the approved shelf arrangement, and R3 for store lighting.

Create a retail discovery shot that makes [PRODUCT] easy to find. Preserve the exact hero SKU from Reference 1. Recreate the approved shelf structure and neighboring category context from Reference 2 without copying unrelated trademarks, and use the store lighting from Reference 3. Begin slightly wide, move forward along one clean path, and settle with the hero product centered and sharp while surrounding packs remain secondary. No invented price tags, promotional claims, duplicate hero packs, changing shelf geometry, shoppers, or generated signage.

This prompt communicates retail context without relying on synthetic offer copy. Add real price and promotion information through the publishing platform, not the generated scene.

3.16 Seasonal Campaign Styling - 4:5

Use R1 for the product, R2 for the approved seasonal palette, and R3 for material cues.

Create a seasonal campaign scene for [PRODUCT] while keeping the package unchanged. Reference 1 is the exact product identity. Use Reference 2 for the approved [SEASON/EVENT] palette and Reference 3 for natural materials such as [MATERIALS]. Begin with a quiet product portrait, introduce gentle environmental motion such as moving light or drifting leaves, then finish on a stable hero frame. Keep the styling contemporary and non-denominational unless the brief requires otherwise. No package recoloring, symbols not present in the references, greeting text, artificial snow on the label, or extra products.

Because the product itself stays fixed, the same master can support multiple seasonal placements. Cultural symbols should always come from an approved reference rather than model improvisation.

3.17 Market Localization Without Generated Text - 9:16

Use R1 for the global product, R2 for the approved local setting, and R3 for local talent or hands.

Create a localized vertical product moment for [MARKET]. Preserve the global product from Reference 1 exactly, including package language already printed on it. Use the approved setting details from Reference 2 and the authorized local talent or hands from Reference 3. Show one everyday action: [ACTION], filmed with warm observational camera movement. Keep product scale, skin tone, clothing, architecture, and objects coherent. Do not generate translations, flags, stereotypes, new labels, extra people, or culturally specific symbols that are not in the references.

This format localizes context rather than inventing localized claims. Real translated copy can remain in the ad or page layer where it is reviewable and accessible.

3.18 Premium Gift Moment - 16:9

Use R1 for the product, R2 for approved gift packaging, and R3 for the recipient's hands and setting.

Create an elegant gifting moment for [PRODUCT]. Preserve the exact product from Reference 1 and the approved gift box, ribbon, and insert from Reference 2. Match the hands and interior setting from Reference 3. Open on the wrapped gift, let the recipient uncover the product in one calm movement, and finish with the product held clearly toward the light. Keep the emotion understated and the object handling realistic. No extra jewelry, altered ribbon colors, duplicate products, card text, face generation, or package deformation.

The shot avoids tying the asset to a single written occasion. It can support birthday, holiday, loyalty, or corporate-gifting copy outside the video.

3.19 Refill or Reuse Demonstration - 4:5

Use R1 for the main container, R2 for the approved refill component, and R3 for the demonstrated action.

Create a clear refill demonstration for [PRODUCT SYSTEM]. Preserve the exact main container from Reference 1 and the exact refill component from Reference 2. Follow the hand position and action order shown in Reference 3. Use one stable three-quarter camera angle as the refill is opened, aligned, and inserted in a physically plausible sequence, then end on the completed product. Maintain correct part count, fit, scale, and materials. No spills, disappearing components, invented sustainability claims, extra hands, altered closures, or instruction text.

The video should show the mechanism, not make an environmental claim. Any quantified sustainability message needs separate evidence and approved copy.

3.20 Campaign Master With Modular End Frame - 16:9

Use R1 for the product, R2 for brand world, R3 for talent, and R4 for the closing composition.

Create a campaign master video for [PRODUCT] that can be adapted across channels. Preserve the exact product from Reference 1, use the lighting and palette from Reference 2, and match the approved talent styling from Reference 3. Begin with a tactile product interaction, transition through one smooth camera arc, and resolve into the clean end-frame composition from Reference 4. Hold the final frame with uncluttered negative space on [LEFT/RIGHT] while keeping the product fully visible. No generated headline, altered package, extra talent, discontinuous props, fast cuts, or watermark.

This is one of the more adaptable prompts in the set because the final composition can carry platform-native copy without baking synthetic text into the video. Protect the clean hold during generation rather than trying to manufacture it later.

4. Turn One Prompt Into a Campaign System

Do not generate all 20 prompts at once. Start with the asset that defines the campaign's visual grammar: usually the launch hero, marketplace turntable, or campaign master. Treat the approved result as a locked brief. Record the exact files assigned to product identity, brand world, talent, and environment; save the accepted prompt, aspect ratio, and motion description; and note which package details passed review. When you branch into variants, duplicate that brief and change only the delivery composition and primary action. Keep the product master and brand-world references unchanged unless the campaign itself has changed. This makes it possible to trace a drifted result back to one deliberate edit instead of several simultaneous variables.

For a practical batch, create a wide master, a vertical social variant, and a square or 4:5 commerce variant. The aspect ratio should change the staging, not merely crop it. In the vertical version, center the action and reserve deliberate space above or below the product for interface elements and approved copy. In the square or 4:5 version, shorten the camera move, keep product scale stable, and bring the feature closer to the center. In the wide version, use environmental context or a measured lateral move without letting the product become too small. Keep the same opening identity check and final product hold across all three versions so reviewers can compare like with like.

Before approving a variant, compare it with the master in the same order every time: silhouette, closure or hardware, label position, product color, relative scale, key lighting, then motion. Reject a visually attractive result if it changes a non-negotiable product detail. Name exports by campaign, format, prompt version, and review state so an approved master cannot be confused with an experiment. When you are ready to test these prompts, open Wan 2.7 in Vofy Studio and confirm the current reference-input and delivery controls. The Wan 2.7 model page provides the model-specific entry point. Vofy is an all-in-one AI creative studio, and this guide uses its workflow as an example; model availability, controls, and Credits can change, with rates varying by model, resolution, duration, and generation mode.

5. Troubleshooting Product Drift and Weak Motion

If the package changes, remove style instructions before adding more constraints. Run a plain identity test with R1, a simple studio background, and a shallow camera move. Once the cap, silhouette, label placement, and material stay stable, reintroduce R2. This isolates whether the conflict comes from the source images or from the motion brief. A highly stylized reference can dominate a weak product packshot, especially when its lighting hides the edges you need Wan to preserve. If the plain test still drifts, replace R1 with a sharper, unobstructed packshot at the intended viewing angle before rewriting the prompt.

If the shot feels static, do not add three new actions. Strengthen one motion verb and give it a visible start and finish: “the hand lifts the bottle from the table and returns it” is easier to stage than “show an energetic lifestyle moment.” If the output feels chaotic, reduce either the camera move or the subject move. A stable camera with one clear product action often creates a more usable ad than a complex orbit combined with flying props and animated light. When motion stops too early, specify the final hold duration and describe the last frame instead of adding another transition.

If labels, reflections, or hands fail, diagnose them separately. For damaged label text, keep the product front-facing, reduce rotation, and shorten the focal transition; legal or ingredient copy still requires frame-by-frame review. For duplicate products or mirrored pumps, remove reflective props and ask for one visible product count. For distorted hands, simplify the grip, keep the wrist visible, and use a talent reference whose perspective matches the product shot. For inconsistent scale, add a familiar surface or approved hand reference rather than relying on adjectives such as “small” or “premium.”

If a vertical output hides the product behind interface elements, regenerate with deliberate safe space rather than planning a rescue crop. Review current placement guidance in the Meta Ads Guide and TikTok Creative Center for the destination you intend to use. For provider-level model updates, consult the official Wan site. These resources change over time, so confirm channel requirements at export and publishing time.

6. Conclusion

The strongest Wan 2.7 multi-reference prompt is not the one with the most cinematic language. It is the one that makes ownership clear: this reference owns product identity, that reference owns the visual world, another owns talent or scale, and the prompt owns motion. Once those responsibilities are explicit, one approved master can become a coherent family of launch, social, commerce, retail, and localized assets. Do not confuse more references with more control; a smaller source pack with non-conflicting roles is usually the better production brief.

FAQ

How many reference images should I use with Wan 2.7?

Use the fewest references needed to define the shot, and confirm the current input limit in the model interface. A practical source pack often includes a product identity image, one brand-world image, and one talent or environment image. More references can help only when each has a distinct role and does not contradict the others.

What if the interface does not label my references as R1, R2, and R3?

Treat R1, R2, and R3 as planning labels for your source pack. Upload the images in the same order used by the prompt, then identify each reference by a visible feature such as “the teal bottle,” “the warm vanity,” or “the hand with a cream sleeve.” If the interface changes the order or does not expose reference IDs, rewrite the role line with those visual descriptions before generating.

Should I put the aspect ratio inside the prompt?

Choose the aspect ratio in the generation controls. You can describe “vertical,” “square,” or “wide” composition in the prompt to guide staging, but prompt wording should not replace the actual output setting. Compose for the final channel from the beginning instead of relying on a crop.

Can Wan 2.7 keep small label text perfectly readable?

Treat exact small typography as a fragile element in moving shots. Use a front-facing product, restrained camera motion, adequate product size, and a stable end frame when readability matters. Review every frame before publication, and do not publish altered legal, ingredient, safety, or claim text.

What should I do when two references have different lighting?

Declare which reference owns the lighting and use the other only for its assigned role. If the product packshot has neutral light and the style board has a strong color cast, preserve product color from the packshot while borrowing the direction and contrast of the style board. If color accuracy still drifts, replace the style reference with a closer match.

Can I use customer, creator, or employee photos as talent references?

Use only images you own or have permission to use, and make sure the intended campaign use is covered. Do not use a real person's appearance to imply an endorsement they did not approve. Keep consent records with the campaign assets and review generated hands, faces, and interactions before publication.

Related Reading

• How to Create Image-to-Video Shots With Kling 3.0
• Seedance 2 Prompt Guide and Playbook
• Kling 3.0 Model Guide