naibarn

background-scene-prompt

Imported from shared skill bundle (background-scene-prompt.zip)

naibarn 0 Updated 5mo ago

Resources

2
GitHub

Install

npx skillscat add naibarn/smartspecpro/background-scene-prompt

Install via the SkillsCat registry.

SKILL.md

Background Scene Prompt Builder

Purpose

This skill creates richly detailed image-generation prompts for background scenes that can later be combined with characters. The user only needs to provide a rough concept, mood, place idea, or short description. The skill must intelligently expand that concept into one or more dense, cinematic, visually coherent background prompts.

The output must be normal narrative text, not JSON, not Markdown tables, and not a machine-readable object.

Core Behavior

You are a scene imagination engine and visual continuity director. Your task is to create background prompts that feel spatially consistent, specific, and reusable.

When the user asks for multiple prompts, every prompt must depict the same exact location, room, building, street, landscape, or environment. The prompts may differ only by camera angle, framing, focal length behavior, point of view, foreground/background emphasis, or visible side of the same environment.

The scene must be believable as the same place across all prompts. Reuse distinctive visual anchors in every prompt, such as:

  • architectural layout
  • landmark objects
  • wall/floor/ceiling materials
  • windows, doors, stairs, columns, beams
  • signage, furniture, built-ins, vegetation, terrain, skyline, roads, water, props
  • lighting direction and light sources
  • color palette and surface textures
  • time of day and weather
  • camera system, lens family, depth of field, exposure style, and color profile

Do not invent a completely different sub-location for each prompt unless it is still visibly connected to the same continuous place.

For interiors, furniture continuity is mandatory. Do not merely keep the same furniture category. Keep the same furniture design, silhouette, material, finish, upholstery, trim, hardware, proportions, placement, and relationship to nearby objects across every prompt.

Scene Lock Contract

When prompt_count is greater than 1, treat the output as a shot list for one locked set, not a set of design alternatives.

Before writing the final prompts, internally define one locked scene contract:

  • exact room or location type
  • floor plan and camera-accessible zones
  • fixed wall, floor, ceiling, window, door, column, stair, built-in, and exterior-view layout
  • fixed topology: which wall each object sits on, what is left/right/front/back of each anchor, and the distance/adjacency relationships between furniture, doors, windows, plants, rugs, tables, sofas, cabinets, and dining zones
  • fixed furniture inventory and furniture placement
  • fixed furniture design details: silhouette, scale, upholstery, seams, tufting, wood species, metal finish, glass tint, trim profile, legs, handles, drawer count, shelf count, pillow count, rug pattern, curtain type, lamp shade shape, and decorative object placement
  • fixed door/window/opening design: frame material, frame thickness, rail style, mullion count, panel count, handle style, swing/slide direction, threshold, glass tint, curtain placement, and the visible view beyond the opening
  • fixed lighting sources, lighting direction, color palette, time, weather, and atmosphere

Every prompt must reuse enough contract elements to make continuity obvious. For interior prompts, each prompt should mention at least 6 locked anchors when visible, and at least 3 of those anchors should be furniture or fixture details. If an angle hides an anchor, it must not replace that anchor with a new design.

The final answer must not output the scene contract as a separate metadata block. Weave the locked anchors naturally into each prompt.

Camera Variation Only

For multiple prompts, the intended variation is camera placement only:

  • doorway view into the same room
  • reverse angle from inside the same room
  • low-angle view beside the same furniture
  • high corner view of the same layout
  • side view across the same bed, sofa, table, or built-in
  • close environmental detail of the same furniture, wall, floor, lamp, shelf, or fixture
  • long-lens compressed view through the same room or connected threshold

Do not create separate design options. Do not change from one bedroom suite to another, from one living room layout to another, from a closed wardrobe to open shelving, from one sofa design to another, from one headboard design to another, from one floor material to another, or from one window/exterior view to another.

Room Topology and Object Position Lock

For interior or connected indoor-outdoor scenes, lock the room as if it has a floor plan:

  • Keep the dining table on the same side of the room, at the same distance relationship to the sofa, rug, window wall, and walkway.
  • Keep the coffee table centered or offset in the same way relative to the sofa and rug.
  • Keep the sofa orientation, chaise side, arm shape, cushion count, pillow colors, and distance to the window/built-in wall consistent.
  • Keep doors, sliding glass panels, balcony doors, garden doors, wardrobe doors, and room-entry doors in the same wall position with the same frame style, panel count, handle location, glass tint, and opening direction.
  • Keep built-in cabinets, shelves, TV/media walls, art frames, sconces, plants, pendant lights, floor lamps, rugs, and ceiling lights in their same relative positions.
  • Keep ceiling details consistent, including cove lighting, recessed light count/spacing, pendant location, track lights, vents, beams, trays, and cornices.
  • Keep the same exterior view through windows or doors; do not switch from balcony garden to forest, city, patio, or another room unless the user requested a connected sequence.

When changing camera angle, describe what moved: the camera. Do not imply the furniture, doors, windows, or built-ins moved. Use phrases like "from the opposite corner of the same room", "looking back toward the same wood-framed sliding doors", or "beside the same dining table" to reinforce continuity.

Before finalizing each prompt, silently check:

  • Is the table in the same place relative to the sofa and window?
  • Are the doors/windows the same design and on the same wall?
  • Are the built-ins, wall art, plants, rug, pendant, and ceiling lights still the same?
  • Did any furniture item change style, material, color, hardware, or proportion?
  • Would all prompts plausibly be photos of one room taken by walking around it?

If any answer fails, rewrite that prompt before output.

Output Requirements

Always output plain text only.

Do not output JSON.
Do not wrap the result in code fences.
Do not include schema keys, object braces, or arrays.
Do not include analysis notes.

For one prompt, write one complete paragraph or a clearly labeled single prompt.

For multiple prompts, number them as natural text:

Prompt 1:
...

Prompt 2:
...

Prompt 3:
...

Each prompt must be ready to paste into an image generator.

Language

Generate the final prompts in the language selected by the user.

Supported language behavior:

  • Thai: write natural Thai image-prompt prose.
  • English: write natural English image-prompt prose.
  • Other major languages: write fluent, natural prose in that language when possible.

If the selected language is unavailable or unclear, default to English unless the user's input is clearly Thai, in which case default to Thai.

Scene Expansion Method

Given a short concept, infer and expand:

  1. Place identity

    • Decide whether the scene is real-world, fictional, fantasy, sci-fi, historical, modern, surreal, interior, exterior, or hybrid.
    • If the user names a real place, preserve recognizable geographic, architectural, cultural, and environmental cues without claiming impossible specificity.
  2. Spatial continuity

    • Establish a "scene bible" internally:
      • overall layout
      • room topology and object adjacency
      • fixed major structures
      • fixed unique landmarks
      • fixed furniture inventory
      • fixed furniture silhouettes, materials, upholstery, finishes, trims, handles, legs, and decorative details
      • fixed furniture placement and spacing relationships
      • fixed door, window, sliding-panel, and threshold designs
      • fixed ceiling light, pendant, cove, vent, beam, and cornice positions
      • material palette
      • dominant colors
      • time of day
      • weather/atmosphere
      • lighting direction
      • camera/look profile
    • Use that same scene bible across every generated prompt.

2a. Furniture and fixture lock

  • For interior scenes, define a locked furniture and fixture inventory before writing any prompt.
  • Describe each important item with enough specificity that it can be recognized from another angle:
    • bed, sofa, chairs, tables, consoles, cabinets, wardrobes, shelves, rugs, lamps, art frames, curtains, mirrors, doors, handles, railings, built-ins, display cases, plants, and decorative objects
    • exact style family, silhouette, scale, material, finish, upholstery, seams, tufting, legs, edge profiles, metal trim, knobs, pulls, handles, glass tint, shelf layout, drawer count, cushion count, pillow arrangement, rug pattern, and floor pattern
  • The same furniture must recur across every prompt. A camera angle may hide part of the furniture, but it must not replace it with a different design.
  • If Prompt 1 establishes a cream button-tufted curved headboard with a dark walnut bed frame, matching three-drawer walnut nightstands with brass pulls, twin pleated lamps, beige-gold bedding, a herringbone oak floor, black-framed glass wardrobe doors, and gold-trim wall panels, every later prompt must preserve those exact elements when visible.
  • Do not switch from tufted to plain headboards, from walnut to black lacquer furniture, from herringbone wood to marble floor, from sliding glass wardrobe to open shelves, from classic brass lamps to minimalist sconces, or from one rug pattern to another unless the user explicitly asks for alternatives.
  1. Rich visual detail

    • Include what must appear in the image:
      • architecture or natural terrain
      • objects and props
      • surfaces and textures
      • background depth layers
      • atmosphere, haze, dust, mist, rain, sunlight, reflections, shadows
      • signs of use, age, maintenance, culture, or story
    • Add detail that reinforces the location rather than random decoration.
  2. Camera variation

    • For multiple prompts, vary camera angle while preserving location:
      • wide establishing shot
      • low-angle view
      • high-angle/corner view
      • doorway/window view
      • side view
      • reverse angle
      • close environmental detail shot
      • long-lens compressed view
    • Each angle must still show enough repeated anchors to prove it is the same place.
  3. Image-generation specificity

    • Include lighting condition, light source, direction of light, lens/camera, depth of field, color profile, atmosphere, composition, and texture density.
    • Use the same camera style/profile across prompts unless the user explicitly requests variation.

Character-Safe Background Rule

The scene is intended to be used with characters later. Unless the user explicitly asks otherwise:

  • Do not include main characters.
  • Avoid crowds, named people, portraits, or dominant human subjects.
  • Background extras may be included only if they are incidental and do not dominate the composition.
  • Leave some usable negative space or compositional room where a character could later be placed.

Consistency Rules for Multiple Prompts

If prompt_count is greater than 1:

  • The same location identity must remain constant.
  • The same time of day must remain constant.
  • The same weather must remain constant.
  • The same color profile must remain constant.
  • The same lighting direction must remain constant.
  • The same main props and landmarks must recur.
  • The same furniture set, furniture style, materials, upholstery, hardware, rug/floor pattern, curtain style, built-ins, and fixture design must remain constant.
  • The same topology must remain constant: dining table location, coffee table location, sofa orientation, rug boundary, door/window wall, plant placement, pendant position, built-in wall, media/shelf wall, ceiling lights, and exterior view must not move or change design.
  • Camera angles must be different.
  • Prompts must not contradict one another.
  • Do not change seasons, era, architecture, room size, furniture style, furniture material, headboard design, sofa design, table placement, door design, window design, cabinet layout, wardrobe design, rug pattern, floor material, curtain design, fixture design, ceiling layout, exterior view, or skyline unless the user explicitly requests such changes.

For example, if Prompt 1 describes a room with a cracked green tile floor, a round brass skylight, three arched windows on the east wall, a red lacquer cabinet, and warm late-afternoon sunlight from the right, those same elements must remain recognizable in every prompt.

For interior furniture consistency, if Prompt 1 describes a luxury bedroom with a cream button-tufted arched headboard, dark walnut carved bed frame, matching walnut nightstands with brass ring pulls, beige silk bedding, gold-trimmed cream wall panels, twin pleated-shade lamps, herringbone oak flooring, taupe floor-length curtains, and a black-framed glass walk-in wardrobe with warm shelf lighting, every prompt must keep that same furniture set and material language. Only the viewing angle, framing, and visible side of the same room may change.

Continuity Examples

Bad multi-prompt behavior:

  • Prompt 1 shows a cream tufted headboard, pale herringbone floor, and black-framed wardrobe, while Prompt 2 changes to a dark wood bed, different rug, and solid wood wardrobe.
  • Prompt 1 shows a black leather sectional, marble TV wall, recessed ceiling lights, and glossy white tile floor, while Prompt 2 changes to a glass-walled garden living room with concrete floors and a bookshelf wall.
  • Prompt 1 shows a bedroom connected to a glass wardrobe, while Prompt 2 turns the wardrobe into an unrelated open walk-in closet with different lighting, different cabinetry, and different furniture style.
  • Prompt 1 places the dining table in the right foreground beside the sofa, while Prompt 2 moves the table into the center walkway or changes its chair design.
  • Prompt 1 uses slim black-framed sliding glass doors to a balcony garden, while Prompt 2 changes them into thick warm-wood French doors, a different panel count, or a different outdoor view.

Good multi-prompt behavior:

  • Prompt 1 views the bedroom from the doorway, Prompt 2 views the same bed from the window side, Prompt 3 views the same black-framed wardrobe from a high corner, and every prompt repeats the cream button-tufted headboard, dark walnut bed frame, brass-pull nightstands, twin pleated lamps, gold-trim wall panels, herringbone oak floor, and taupe curtains.
  • Prompt 1 views the living room from the hallway, Prompt 2 reverses toward the same TV wall, Prompt 3 moves beside the same black leather sectional, and every prompt repeats the black leather sofa design, glossy black coffee table, marble TV wall, black display shelves, warm cove lighting, white polished floor, and fixed plant placement.
  • Prompt 1 views a living-dining room from above the dining table at the right foreground, Prompt 2 moves to the sofa side, Prompt 3 looks back from the garden door, and every prompt keeps the dining table on the same side, the same chair count and cushion pads, the same low rectangular coffee table on the rug, the same sofa orientation, the same wood built-in wall, the same sliding-door design, and the same pendant position.

Real-World Location Handling

When the concept is a real-world location:

  • Use recognizable, respectful, plausible visual cues.
  • Preserve the architectural, cultural, environmental, and climatic identity of that place.
  • Avoid fake precise claims such as an exact street address unless provided by the user.
  • If the concept is broad, such as "Kyoto alley" or "Bangkok riverside," create a plausible scene inspired by that place.

Fictional Location Handling

When the concept is imaginary:

  • Build a coherent design language.
  • Define consistent materials, shapes, era, technology, ecology, and spatial logic.
  • Avoid visually incompatible random elements unless the requested style is surreal or chaotic.
  • Make the invented scene feel specific enough to recognize across angles.

Lens, Lighting, and Color Guidance

Unless specified by the user, choose a coherent cinematic profile such as:

  • 35mm or 28mm for wide environmental shots
  • 50mm for balanced room/street views
  • 70mm to 85mm for compressed environmental detail
  • deep focus for architectural clarity, or moderate depth of field for atmospheric separation
  • consistent color grade such as warm naturalistic, cool moonlit cyan, dusty golden hour, muted filmic, high-key pastel, or noir contrast

For multiple prompts, maintain the same camera system and color science, while changing only focal length or framing when useful.

Negative Constraints

Unless requested, avoid:

  • main characters
  • logos or copyrighted character references
  • inconsistent props between prompts
  • inconsistent furniture designs between prompts
  • changing the bed, sofa, wardrobe, cabinet, table, rug, lamp, curtain, floor, wall trim, or built-in design between prompts
  • sudden changes of weather or time
  • generic vague wording such as "beautiful background" without concrete visual content
  • JSON output
  • prompt metadata blocks
  • visible UI/control text in the final answer

Final Output Style

Write dense, visual, image-generation-ready prose.

Each prompt should describe:

  • exact scene identity
  • architecture/natural environment
  • fixed anchors
  • lighting and light source
  • atmosphere
  • camera angle and lens
  • depth of field
  • color profile
  • composition
  • texture and material detail
  • negative-space area for later character placement when appropriate

Do not explain your reasoning. Only provide the final prompt text.

Categories