| name | hbg-oriental-giant-world |
| description | Convert a keyword, rough idea, existing prompt, or reference-derived scene into a luminous Eastern colossal world with an ontologically impossible giant construct, strong cognitive contrast, cinematic stills, and image-relative Veo/Gemini motion. Use for 东方美学、东方巨物、万物皆可巨构、荒诞尺度、仙侠神话、文化题材东方化或统一图片与视频风格。 |
HBG Oriental Giant World
Transform subject matter into the HBG Eastern Colossal Style Lock without copying a reference composition or erasing the source culture.
Read first
- Read
references/oriental-style-lock.md completely.
- Read
references/prompt-contract.md before writing the still and motion prompts.
- For image generation, use the environment's image-generation skill and obey visible-delivery rules.
- For video creation, read the environment's mandatory video skill first. Keep Gemini/Veo when explicitly requested.
Accepted inputs
- one keyword:
道教, 印度尼西亚, 月球, 弹琴;
- a rough scene idea;
- an existing image prompt;
- a reference image or video;
- a prompt-bank question selected by a roaming skill.
If a reference is supplied, extract only transferable style signals: material precision, atmosphere, light, palette, lens energy, and motion rhythm. Do not inherit its circular gate, centered axis, central figure, symmetric courtyard, tree frame, or any other composition skeleton unless the user explicitly requests that composition.
Transformation contract
Preserve:
- the source subject, action, cultural identity, symbolic meaning, and core wonder;
- culturally specific architecture, ritual boundaries, clothing, ecology, and materials;
- one memorable semantic hook from the source.
Transform:
- the subject itself into terrain, architecture, weather, a world-body, or an impossible spatial law;
- spatial design into a physically rendered but conceptually impossible multi-layer colossal world;
- lighting into luminous, airy, auspicious, high-clarity cinema;
- scale into tiny-human-versus-inhabited-world-body contrast with at least three scale proofs;
- composition into a new, non-template wide-angle arrangement;
- motion into slow, image-relative, rigid-anchor Veo language.
Do not “Chinese-wash” Thailand, Indonesia, Hinduism, Buddhism, or other living cultures. The target Eastern aesthetic is a spatial, material, tonal, and cinematic grammar; regional identity remains dominant. Use fully Chinese mythic architecture only when the subject supports it.
Absurdity gate
Reject a design if it is only a real temple beside a volcano, a normal sea gate, a village under an enlarged roof, a palace in clouds, or another plausible tourist landscape. “Large” is not enough.
Require all of these:
- the subject itself changes category;
- at least two domains fuse inseparably, such as anatomy + city, creature + sky, craft + geology, object + continent, or action + weather;
- one impossible law produces visible consequences in water, cloud, light, gravity, horizon, or habitation;
- tiny people visibly cross, inhabit, excavate, maintain, or worship within the giant construct;
- at least three independent scale proofs remain readable.
Keep materials and light realistic. Keep the premise impossible.
Style lock
Unless the user explicitly asks for darkness or horror:
- luminous, airy, auspicious, serene, sacred, inviting;
- readable high-key image with blue-white atmosphere and warm dawn or dusk gold;
- ivory stone, pale jade, aged timber, vermilion, bronze, restrained gold, regional materials;
- realistic air perspective, volumetric clouds or mist, water vapor, and long sightlines;
- one giant form, one atmospheric or celestial anchor, tiny people, generous negative space;
- Chinese mythic creatures must be slender, elegant, auspicious, and culturally recognizable—not Western attacking dragons.
Avoid default storm-grey grading, crushing black foregrounds, horror faces, apocalyptic debris, neon sci-fi, generic European castles, Japanese shrine substitution, visual clutter, and character close-ups.
Composition anti-template gate
Before each new design, name the previous foreground family, camera relationship, subject position, and leading-line geometry. Change at least three of the four.
Rotate among:
- monumental gate, eave, colonnade, wall opening;
- canyon, cave, cliff, volcanic shelf, ice wall;
- water plane, waterfall curtain, rain veil, boat bow;
- statue base, relief wall, ceremonial vessel, giant bridge;
- open ground with shadow or extreme proximity;
- vegetation only when semantically justified.
Support low, eye-level, high-oblique, aerial, and selective top-down views. A batch of four should feel like four discoveries in one world, not four skins on one layout.
Still workflow
- Produce an adaptation card: source semantics, reality baseline to reject, giant transmutation, category fusion, impossible law, visible consequences, inhabitation, three scale proofs, camera, foreground, atmosphere, materials, light, and prohibitions.
- Write the exact 16:9 still prompt using the Style Lock.
- Reject the concept before generation if it could exist as a normal travel photograph.
- If requested, generate and inspect the mother image.
- Reject reference-composition copying, weak giant dominance, missing inhabitation, fewer than three scale proofs, malformed architecture, fake writing, duplicated people, generic fantasy substitution, over-darkness, and collapsed depth.
- Deliver only after the image is visibly present in chat and saved to an absolute path.
Motion workflow
Write from the actual approved image. Choose exactly one main camera path: stabilized dolly, crane rise, lateral arc, descending reveal, or slow axial passage. Add restrained environment motion already supported by the image: layered cloud drift, waterfall spray, fabric, incense, foliage, water reflection, dust, rain, or light breathing.
Architecture, terrain, deity form, planet, moon, bridge, and other giant anchors remain rigid. People perform at most one small plausible action. No new buildings, creatures, symbols, crowds, readable text, energy beams, weather replacement, scene cut, time-lapse, fast orbit, or camera collision.
Output contract
Always return the adaptation card, complete still prompt, complete image-relative motion prompt, and explicit constraints. When generating media, also return the visible still, saved files, verified video specifications, and QA notes.
