Quick Answer

Realistic AI product photos need separate instructions for photographic feel, subject arrangement, product identity, color, exclusions, and reuse. The six patterns are Subject + Surface + Light + Lens + Finish; Framing or Pose + Camera View; Reference-Image Anchoring; Color Lock; Constraint and Exclusion; and Save and Apply. Each pattern controls one part of the result, which makes a failed image easier to diagnose and a successful setup easier to reuse.

Why use prompt patterns instead of copy-paste prompts?

A prompt pattern is a repeatable set of slots, while a copy-paste prompt is one finished instruction for one image. Patterns let you change the product without accidentally changing the lighting, or change the crop without rewriting the product description.

The six patterns below preserve the useful parts of a successful brief as separate decisions. They can be combined, but they do not depend on one another. That separation matters because the source of a bad result becomes visible: product identity belongs in the reference pattern, camera geometry belongs in the arrangement pattern, and unwanted elements belong in the constraint pattern.

Different image systems expose different controls. Stability AI's image API exposes a dedicated negative_prompt parameter, while Google's Gemini image API uses natural-language instructions and can accept multiple images as context. Production systems may store the same decisions as reusable controls. Treat each pattern as an intent, then express it through the interface the system actually provides.

For the broader production problem, read the guide to consistent AI product photography.

Pattern 1: Subject + Surface + Light + Lens + Finish

Subject + Surface + Light + Lens + Finish gives the model a compact description of what a plausible product photograph should look like. It is a useful starting point when an image feels synthetic because the material, lighting, depth of field, or finish is underspecified.

What is the Subject + Surface + Light + Lens + Finish slot template?

The Subject + Surface + Light + Lens + Finish template separates five visual decisions so each one can be changed without disturbing the others.

[Subject with material and color] on [surface or background], [light source and direction], [camera and lens cue], [finish and quality terms]

What does a complete Subject + Surface + Light + Lens + Finish prompt look like?

A complete Subject + Surface + Light + Lens + Finish prompt names one clear choice for each slot.

A matte ceramic coffee mug in warm cream,
on a smooth concrete surface,
lit by one large softbox at 45 degrees from camera left,
photographed with a 100mm macro lens at f/8,
sharp commercial product photography with fine surface detail
and a soft directional shadow.

Camera and lens terms act as visual cues, not as a physical camera simulator. A focal length and aperture can suggest perspective and depth of field, but the generated image still needs review. The same restraint applies to lighting: one source with a direction and softness is usually easier for a model to interpret than several conflicting setups.

This pattern controls material reading, light direction, depth of field, perceived sharpness, and photographic finish. It does not preserve the identity of a real product, fix the same crop across a catalog, or guarantee an exact output color.

What commonly goes wrong with Subject + Surface + Light + Lens + Finish prompts?

Subject + Surface + Light + Lens + Finish prompts usually fail when several lenses, lighting setups, or finish adjectives give the system conflicting direction. Keep one deliberate choice per slot, generate, and revise the slot responsible for the visible problem.

How does Nightjar make the photographic feel reusable?

Nightjar stores the photographic language of a shoot as a reusable Photography Style. A Photography Style controls camera feel, lighting, mood, color scheme, texture, and atmosphere, while the surface or scene remains a separate Background choice and the arrangement remains a separate Framing or Pose choice.

Teams can choose from 150+ curated Photography Styles or create a custom Photography Style from exactly three reference Assets. Reusing that Photography Style makes it easier to continue the same visual direction across later Products and Generations without rewriting lens-and-light language. The Photography Style guide explains the separation in more depth.

Pattern 2: Framing or Pose + Camera View

Framing or Pose + Camera View controls geometry without mixing it into lighting or mood. Product-only shots need instructions for product staging and crop; images with a person need instructions for body arrangement and camera distance.

What is the Framing or Pose + Camera View slot template?

The Framing or Pose + Camera View template changes according to whether the product appears alone or with a person.

Product-only: [shot type and angle] of [product], [staging], [crop and scale], [negative space]

With a person: [body arrangement], [how the product is worn or held], [camera distance], [camera angle and crop]

What does a product-only Framing + Camera View prompt look like?

A product-only Framing + Camera View prompt should make the camera angle and the product's size in the frame unambiguous.

Three-quarter product shot of the bag,
rotated about 30 degrees from front,
camera slightly above eye level,
product centered and filling most of the frame,
with clear negative space above, 4:5 aspect ratio.

Arrangement terms describe geometry, not the photographic finish. Keep “three-quarter product shot” separate from “softbox at 45 degrees,” because the first describes the camera and subject while the second describes the light.

This pattern controls angle, staging, body arrangement, crop, product scale, and negative space. It does not control the setting, lighting, material, or product identity.

What commonly goes wrong with Framing or Pose + Camera View prompts?

Framing or Pose + Camera View prompts become ambiguous when they use “angle” without naming its subject. State “camera slightly above the product” for camera position and “light at 45 degrees from camera left” for illumination.

How does Nightjar separate product-only and on-model arrangement?

Nightjar uses a Framing setting for product-only shots and a reusable Pose plus Camera Distance for shots with a Fashion Model. Framing selects camera angle, staging, and crop from a fixed visual set; Pose stores body arrangement, while Camera Distance selects a close, medium, or full-body view supported by that Pose.

Composition is retired Nightjar terminology. The current separation prevents a product-only camera angle from being confused with a Fashion Model's pose or with the Background behind the subject. The camera-angle guide covers the product-only controls.

Pattern 3: Reference-Image Anchoring

Reference-Image Anchoring gives the system visible evidence for a product, scene, person, or photographic direction that text alone cannot specify reliably. The instruction should assign one explicit role to each reference.

What is the Reference-Image Anchoring slot template?

The Reference-Image Anchoring template names both the reference and the aspect it governs.

Use [reference A] for [product identity or structure]. Use [reference B] for [lighting, setting, person, or style]. Create [new output] while preserving the assigned role of each reference.

What does a multi-reference product-image prompt look like?

A useful multi-reference product-image prompt states which image is authoritative for the real product.

Use @image1 as the actual product and preserve its visible shape,
hardware, stitching, and label design.
Use @image2 only for the lighting and color treatment.
Place the product on a polished walnut surface,
in a three-quarter view with soft window light from camera left,
at a 4:5 aspect ratio.

The reference is evidence, while the prompt is a role assignment. Google's Gemini image guide describes the raw capability plainly: “Provide multiple images as context to create a new, composite scene.” Without explicit roles, those images may provide competing cues for structure, color, setting, and style. Reference anchoring helps with product continuity and style transfer, but it cannot reveal a detail that is absent or obscured in every source image.

What commonly goes wrong with Reference-Image Anchoring?

Reference-Image Anchoring fails when several references arrive without roles and the system must infer which details matter. Use fewer, clearer references and say which one governs the product, setting, person, or photographic treatment.

How does Nightjar preserve reference roles?

Nightjar's Edit tab supports up to eight input Assets and direct @image1, @image2, and similar references inside the edit instruction. That makes “use this Asset for the product and that Asset for the setting” readable and explicit.

For repeat work, a Nightjar Product groups multiple Product Photos, plus an optional factual description and physical dimensions, around one visually distinct sellable item. Richer product evidence gives the Generation more context than one loose image, and built-in visual review can catch and retry obvious eligible failures at no extra Credit cost. This is protection, not a guarantee of perfect logos, text, color, or structure. See the guides to realistic product placement and matching AI photos to existing brand photography.

Pattern 4: Color Lock

Color Lock combines a precise target with a source-preservation instruction. A hex, Pantone, or other color specification narrows the requested range, but the result still needs review under the generated lighting.

What is the Color Lock slot template for AI product photos?

The Color Lock template distinguishes the requested color from the product details that should remain stable.

[Source product], target color [specification]. Preserve [shape, construction, texture, hardware, folds, shadows, and marks]. Keep [lighting and framing] unchanged where possible.

What does a Color Lock prompt for an AI product photo look like?

A useful Color Lock prompt pairs the target value with named preservation priorities.

Recolor the bag in @image1 toward hex #2A6F4A, a deep forest green.
Preserve its shape, stitching, hardware, fold geometry, material texture,
and visible label design. Keep the source lighting and framing unchanged.

Named colors cover broad visual ranges, so a numeric or industry color reference gives clearer direction. It does not guarantee that every output pixel will match the target: illumination, reflections, material response, display conditions, and generation variance all affect perceived color.

This pattern controls the requested color and states which source details matter. It does not replace color-critical review or determine the lighting setup.

What commonly goes wrong with Color Lock prompts?

Color Lock prompts become unreliable when a hex value is supplied without a source or preservation clause and the system regenerates the product around the new color. Anchor the real product, name the construction details that matter, and inspect the output before publishing it as a sellable colorway.

How does Nightjar handle explicit color direction?

Nightjar's Edit tab provides a /color command and a Recolor Edit Shortcut, so the target color and source Asset remain explicit. Nightjar is designed to help preserve lighting, shadows, texture, material cues, and product structure during recoloring, but generated color accuracy still requires review.

Pattern 5: Constraint and Exclusion

Constraint and Exclusion works only when it matches the interface the image system actually exposes. Use a separate negative field when the interface provides one; otherwise put concise positive conditions and necessary exclusions in the main instruction.

What is the Constraint and Exclusion slot template?

The Constraint and Exclusion template has separate negative-field and natural-language forms because the two interfaces express the same intent through different controls.

Negative-field interface: Positive: [desired image] and Negative: [short list of unwanted features]

Natural-language interface: [desired image]. The frame contains [positive description]. Exclude [short list of critical unwanted elements].

What do negative-field and natural-language exclusions look like?

Negative-field and natural-language exclusions should both keep the desired image primary and the exclusion list short.

Positive: studio product shot of a glass perfume bottle on a white background,
single softbox above, 85mm lens at f/8, sharp commercial focus
Negative: duplicate bottle, hands, overlay text, watermark,
visible studio equipment, blur

A studio product shot of a glass perfume bottle on a pure white background.
Use one softbox above and an 85mm lens cue at f/8. The frame contains only
the bottle and its contact shadow. Exclude hands, overlay text, watermarks,
duplicate bottles, and visible studio equipment.

A dedicated negative field and an inline instruction are different contracts. Stability AI documents negative_prompt as an advanced API parameter for concepts that should not appear, while the Gemini image guide demonstrates ordinary natural-language instructions instead. Do not assume a field exists because another image system has one, and do not paste a long keyword blacklist into ordinary prose. Positive phrasing is useful when it can describe a complete desired state, such as “an empty table” or “only the bottle and its shadow.” Direct exclusions remain useful for critical elements that have no clean positive substitute.

This pattern reduces common unwanted elements. It cannot correct weak product references, contradictory direction, or a capability the underlying image system does not have.

What commonly goes wrong with Constraint and Exclusion prompts?

Constraint and Exclusion prompts often fail when a long blacklist dominates the instruction and repeats the concepts the system should avoid. Keep the list short, remove conflicts, and fix the positive description before adding more negatives.

How does Nightjar separate defaults from exceptions?

Nightjar uses structured product-photography controls for the recurring brief and Custom Directions for exceptions. A Product supplies richer subject evidence, the Background choice defines an automatic setting, flat color, Backdrop, or Location, and built-in visual review adds a chance to catch obvious product substitutions, omissions, broken readable marks, or catastrophic defects before completion.

Nightjar's Upscale Workflow is preservation-first and targets a 2K or 4K long edge without creative reinterpretation. It is designed to preserve product content, identity, color, text, logos, and structure, but it should not be described as a guarantee that new details can never appear. The Upscale and product-preservation guide explains the intended use.

Pattern 6: Save and Apply

Save and Apply turns a working set of prompt decisions into a reusable production setup. It prevents small wording and setting changes from accumulating when several Products, people, or campaigns use the same direction.

What is the Save and Apply slot template?

The Save and Apply template preserves how the product should be photographed while leaving the product itself replaceable.

Define patterns 1 through 5 once. Save the direction and output settings as a named setup. Apply that setup to the next product, then review the result.

What should a saved AI product-photography setup contain?

A useful saved AI product-photography setup keeps each production decision in its own field.

• Photography Style: clean studio
• Model inclusion: none
• Framing: three-quarter
• Shadow: soft
• Background choice: solid white
• Custom Directions: “Preserve visible label design and exclude studio equipment from the frame.”
• Aspect ratio: 1:1
• Resolution: 2K
• Output format: JPEG
• Image count: 4 independent shots

Saving the structure is more reliable than saving one long paragraph because each axis remains editable. A producer can change Framing without rewriting lighting, or change the output ratio without touching product-preservation instructions.

This pattern improves repeatability across Products, Team members, and later campaigns. It does not make a weak setup good, guarantee identical outputs, or remove the need for review.

What commonly goes wrong when teams reuse AI product-photo prompts?

Teams create prompt drift when they treat a successful AI product-photo prompt as informal knowledge that someone later retypes or “improves.” Store the approved setup, give it a clear name, and change it deliberately rather than through copied prose.

How do Recipes preserve production direction in Nightjar?

Nightjar has a feature called Recipes: Team-owned reusable Create-form setups that save model inclusion, Photography Style, Framing and Shadow or Pose and Camera Distance, Fashion Model, Background choice, Custom Directions, image count, aspect ratio, resolution, and output format. A Recipe does not save Products, Additional Photos, or generated Assets.

That boundary is the point: Products remember what you are photographing, while Recipes remember how you photograph it. A Team can keep up to 100 active Recipes, and every Team member can reuse the same setup. The consistent AI product photos guide and consistent lookbook guide cover two practical applications.

How do the six prompt patterns work together?

The six prompt patterns work best as separate layers in one brief, because each layer has one job and one owner.

The table also exposes conflicts. If a Photography Style carries one mood while Custom Directions demand another, the brief needs repair. If a prompt requests an overhead product-only shot while an on-model Pose implies a full-body view, choose which subject arrangement the image actually needs.

Where do prompt-only workflows stop scaling?

Prompt-only workflows become fragile when the approved direction must survive handoffs, many Products, or a return to the campaign months later. The problem is no longer prompt quality; it is whether the subject evidence, visual choices, and output settings remain findable and reusable.

Three forms of drift are common:

1. Operator drift: different people paraphrase the same brief and change its meaning.
2. Control drift: style, arrangement, background, and output settings are mixed in one text block, so one edit changes several axes.
3. Reference drift: product and brand references become detached from the instructions that explain their roles.

The structural answer is to store the subject and production direction separately. In Nightjar, a Product keeps its Product Photos, factual description, and optional dimensions together; reusable Photography Styles, Backgrounds, Poses, and Fashion Models store visual direction; Recipes package the Create-form choices; and the Team Library keeps those resources available to collaborators. New Generations can then begin from the same evidence and setup instead of a reconstructed prompt.

Frequently Asked Questions

These answers summarize how to choose, combine, and review the six patterns.

What is the best prompt structure for realistic AI product photos? Use separate slots for the subject, surface or setting, light, lens cue, photographic finish, arrangement, references, color, and exclusions. Keeping those decisions separate makes the prompt easier to diagnose and reuse than one paragraph of adjectives.

Why do AI product photos sometimes look fake or plastic? The prompt may lack material, light-direction, depth-of-field, and finish cues, or the source may not show enough real product detail. Add one deliberate choice for each Pattern 1 slot, then use clear Product Photos or references to ground the product itself.

How do I keep the same product across several AI shots? Use the real product as an assigned reference rather than relying on a text description, and provide multiple useful views when the system supports them. Keep the surrounding direction separate so lighting or background changes do not silently redefine the product.

Can a hex code guarantee an exact product color? No. A hex value gives precise direction, but generated lighting, reflections, material response, and model variance can change perceived color. Pair the value with a source-preservation instruction and review color-critical outputs before use.

Do negative prompts work in every AI image system? No single negative-prompt format works everywhere. Use a dedicated negative field only when the interface provides one; otherwise describe the desired state positively and add a short list of essential exclusions in the main instruction.

How long should an AI product-photo prompt be? The prompt should be long enough to give one clear choice for every relevant slot and no longer. A reference-led edit may need only a few sentences, while a new scene may need separate clauses for the product, setting, light, arrangement, and exclusions.

Can I use a reference image instead of a detailed prompt? Yes, but a reference still needs a role. State whether it governs product identity, structure, person, setting, lighting, or style so the system does not have to infer which part to copy.

How do I reuse a successful prompt across a catalog? Save the photographic direction, arrangement, setting, exceptions, and output settings as separate fields, then swap only the product evidence. In Nightjar, Products hold what is being photographed and Recipes hold how it should be photographed.

References

• Google Gemini API image-generation guide - Natural-language image generation, editing, multiple-image context, and batch generation
• Stability AI image API reference - Dedicated prompt, negative-prompt, image-reference, strength, and output controls
• OpenAI image-generation guide - General image API generation, editing, reference-image, output, and quality controls