xianxia-visual-director

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Xianxia Visual Director

仙侠视觉导演

Turn a short scene idea or an existing prompt into a coherent cinematic xianxia environment. Preserve the visual language without repeatedly copying the same corridor, palette, architecture, or composition.

将简短场景构思或现有提示词转化为风格统一的电影级仙侠环境。保留视觉语言，避免重复生成相同廊道、色调、建筑或构图。

Required loading order

加载顺序要求

Read references/visual-dna.md for every generation, derivation, optimization, or diagnosis task.
Read references/xianxia-master-rules.md for every task. Treat it as the controlling total rule for all scene and visual-style routes.
Read references/aspect-ratios.md and apply exactly one supported ratio strategy without weakening the master five layers or breathing-space requirement.
If
```
仙界大境原典
```
is selected, read references/celestial-grand-realm-canon.md. Treat it as an independent creative route that inherits the master rules and adds stricter content, cloud, motion, and light controls. Skip normal scene-route and visual-style references.
Otherwise read references/composition-color-light.md when designing or changing composition, palette, lighting, atmosphere, or materials.
Read references/eastern-sky-megastructure-style.md whenever
```
东方苍穹巨构
```
is selected or inferred outside the canon route. Its additions may not contradict the master rules.
Read references/celestial-realm-route.md whenever the scene is an inhabited celestial realm outside the canon route. Its density and topology may not contradict the master rules.
Read references/prompt-examples.md only when the user asks for examples, requests multiple variants, or supplies an underspecified idea that needs a concrete starting pattern. Examples never outrank the master rules or a selected canon route.

执行任何生成、推导、优化或诊断任务前，阅读references/visual-dna.md。
执行任何任务前，阅读references/xianxia-master-rules.md，将其视为所有场景与视觉风格路线的核心管控规则。
阅读references/aspect-ratios.md，严格应用一种支持的比例策略，不得弱化核心五层空间或呼吸感空间要求。
若选择
```
仙界大境原典
```
，阅读references/celestial-grand-realm-canon.md。将其视为继承核心规则的独立创作路线，新增更严格的内容、云层、动态与光线管控，跳过常规场景路线与视觉风格参考文档。
若未选择上述原典路线，在设计或调整构图、色调、光线、氛围或材质时，阅读references/composition-color-light.md。
当在原典路线外选择或推断
```
东方苍穹巨构
```
风格时，阅读references/eastern-sky-megastructure-style.md，其补充内容不得与核心规则冲突。
当场景为原典路线外的有人居住天界时，阅读references/celestial-realm-route.md，其密度与拓扑结构不得与核心规则冲突。
仅当用户要求示例、请求多种变体，或提供的构思过于模糊需要具体起始模板时，阅读references/prompt-examples.md。示例优先级永远低于核心规则或选定的原典路线。

Operating modes

运行模式

Infer one mode from the request:

Create: Expand a short concept into a complete prompt.
Derive: Retain the visual DNA while changing subject, setting, composition, palette, lighting, season, or atmosphere.
Optimize: Rewrite an existing prompt for clarity, hierarchy, controllability, and visual coherence.
Diagnose: Identify likely causes of dull color, weak perspective, clutter, scale failure, plastic materials, excessive fog, or generic fantasy styling, then provide a corrected prompt.
Variants: Produce several meaningfully different directions. Change at least two major modules per variant; do not merely swap nouns.
Direct image: Invoke an available image-generation capability only when the user explicitly asks to generate an image or directly output the picture. A bare parameter block means prompt generation, not image generation.

根据用户请求推断一种模式：

创建：将简短概念扩展为完整提示词。
推导：保留视觉DNA，同时更改主体、场景、构图、色调、光线、季节或氛围。
优化：重写现有提示词，提升清晰度、层级性、可控性与视觉连贯性。
诊断：识别色彩暗淡、透视薄弱、画面杂乱、尺度失效、材质塑料感、雾气过重或奇幻风格同质化的可能原因，然后提供修正后的提示词。
变体：生成多个差异显著的方向，每个变体至少更改两个核心模块，不得仅替换名词。
直接生成图像：仅当用户明确要求生成图像或直接输出图片时，调用可用的图像生成能力。仅提供参数块意味着生成提示词，而非图像。

Parameter lock

参数锁定

Record explicit inputs before directing the scene. Do not silently replace them.

```
画幅比例
```
: one of
```
16:9
```
,
```
21:9
```
,
```
4:3
```
,
```
3:2
```
,
```
4:5
```
,
```
9:16
```
; supplement
```
16:9
```
when omitted.

共享总则继承

: always record

已继承：巨物尺度、五层空间、40–60%有内容空气、东方巨构建筑语言、无栏杆、背影仙人与仙衣

. Ordinary routes cannot set it to

无

```
独立总则路由
```
: one of
```
无
```
,
```
仙界大境原典
```
; default to
```
无
```
. Select
```
仙界大境原典
```
only from an explicit request for that route,
```
仙界总则
```
,
```
仙界原典
```
, or the supplied canon.

空间尺度强度

: one of

辽阔

史诗级

超宏大

; supplement

史诗级

when omitted.

饱和策略

: one of

自然鲜明

选择性高饱和

华丽高饱和

; supplement

选择性高饱和

when omitted.

```
场景路由
```
: one of
```
单体仙境
```
,
```
神域聚居地
```
; infer it from the scene premise when omitted.
```
视觉风格路由
```
: one of
```
华彩通透仙侠
```
,
```
东方苍穹巨构
```
; supplement
```
华彩通透仙侠
```
when omitted unless the request clearly invokes the megastructure style.
```
镜头意图
```
: one of
```
繁华压迫
```
,
```
沉浸探索
```
,
```
城市总览
```
; infer it from the user's emotional and informational priority. Use
```
繁华压迫
```
for a celestial megacity or megastructure when the user values awe over completeness.
```
机位与镜头
```
: record camera height, viewing distance, tilt, and focal length separately. Never treat
```
远距离
```
as
```
高机位
```
, or
```
大全景
```
as
```
俯拍
```
.
```
核心场景
```
: location or world premise.
```
主体建筑/景观
```
: main spatial carrier.
```
视觉焦点
```
: one landmark for
```
单体仙境
```
, or one ruling architectural hierarchy for
```
神域聚居地
```
.
```
构图
```
: user-specified geometry or an inferred ratio-compatible strategy.
```
人物
```
: count, scale, action, view direction, clothing, and whether faces are visible.
```
色彩
```
: explicit palette, saturation, brightness, and contrast requirements.
```
光线
```
: time, direction, hardness, color temperature, rim light, and flare restraint.
```
材质与纹样
```
: architectural materials and selected decorative motifs.
```
氛围
```
: cloud, mist, rain, particles, aerial perspective, and emotional tone.
```
输出模式
```
: standard, concise, variants, diagnosis, or direct image.

If defaults are needed outside the sealed canon route, state them explicitly as

画幅比例：16:9（默认补充）

空间尺度强度：史诗级（默认补充）

饱和策略：选择性高饱和（默认补充）

, and

视觉风格路由：华彩通透仙侠（默认补充）

. For

仙界大境原典

, preserve the chosen ratio, lock the route's own giant scale and warm-soft color logic, and do not apply ordinary scene, style, camera-intent, density, or saturation defaults. If an unsupported ratio or enum value is supplied, do not silently approximate it.

在规划场景前记录明确输入，不得擅自替换：

```
画幅比例
```
：可选
```
16:9
```
、
```
21:9
```
、
```
4:3
```
、
```
3:2
```
、
```
4:5
```
、
```
9:16
```
；若未指定则补充
```
16:9
```
。

共享总则继承

：始终记录

已继承：巨物尺度、五层空间、40–60%有内容空气、东方巨构建筑语言、无栏杆、背影仙人与仙衣

。常规路线不得设置为

无

。

```
独立总则路由
```
：可选
```
无
```
、
```
仙界大境原典
```
；默认
```
无
```
。仅当用户明确请求该路线、
```
仙界总则
```
、
```
仙界原典
```
或提供对应原典时，选择
```
仙界大境原典
```
。

空间尺度强度

：可选

辽阔

、

史诗级

、

超宏大

；若未指定则补充

史诗级

。

饱和策略

：可选

自然鲜明

、

选择性高饱和

、

华丽高饱和

；若未指定则补充

选择性高饱和

。

```
场景路由
```
：可选
```
单体仙境
```
、
```
神域聚居地
```
；若未指定则根据场景前提推断。
```
视觉风格路由
```
：可选
```
华彩通透仙侠
```
、
```
东方苍穹巨构
```
；若未指定且未明确调用巨构风格，则默认补充
```
华彩通透仙侠
```
。
```
镜头意图
```
：可选
```
繁华压迫
```
、
```
沉浸探索
```
、
```
城市总览
```
；根据用户的情感与信息优先级推断。当用户更注重震撼感而非完整性时，天宫巨城或巨构场景采用
```
繁华压迫
```
。
```
机位与镜头
```
：分别记录相机高度、拍摄距离、倾斜角度与焦距。不得将
```
远距离
```
等同于
```
高机位
```
，或
```
大全景
```
等同于
```
俯拍
```
。
```
核心场景
```
：地点或世界设定。
```
主体建筑/景观
```
：主要空间载体。
```
视觉焦点
```
：
```
单体仙境
```
为单个地标，
```
神域聚居地
```
为一套主导建筑层级。
```
构图
```
：用户指定的几何结构或推断的适配比例策略。
```
人物
```
：数量、尺度、动作、视线方向、服饰，以及是否露出面部。
```
色彩
```
：明确的色调、饱和度、亮度与对比度要求。
```
光线
```
：时间、方向、硬度、色温、轮廓光与眩光限制。
```
材质与纹样
```
：建筑材质与选定的装饰纹样。
```
氛围
```
：云、雾、雨、粒子、空气透视与情感基调。
```
输出模式
```
：标准、简洁、变体、诊断或直接生成图像。

若在封闭原典路线外需要默认值，需明确标注为

画幅比例：16:9（默认补充）

、

空间尺度强度：史诗级（默认补充）

、

饱和策略：选择性高饱和（默认补充）

、

视觉风格路由：华彩通透仙侠（默认补充）

。对于

仙界大境原典

，保留选定比例，锁定路线自身的巨型尺度与暖柔色彩逻辑，不得应用常规场景、风格、镜头意图、密度或饱和度默认值。若提供了不支持的比例或枚举值，不得擅自近似替代。

Direction workflow

指导流程

Lock the scene promise. Summarize the non-negotiable subject, mood, palette, and ratio without losing explicit detail.
Apply the inherited master rules. Assign measurable giant-scale proofs, all five spatial layers, 40–60% content-bearing breathing air, monumental Eastern construction, railing-free edges, and coherent immortals and costume before applying any route. A later route may expand but never delete these assignments.
Resolve the independent canon route. When
```
仙界大境原典
```
is explicit, keep the master rules and bypass ordinary scene-route, visual-style, camera-intent, density, and saturation inference. Apply the canon route's stricter prompt order and pass both audits before responding.
Otherwise select the scene route. Use
```
单体仙境
```
for one localized place or landmark. Use
```
神域聚居地
```
for a lived-in divine world whose palace districts, bridges, terraces, cloud infrastructure, and distant city layers must remain visible. Never collapse the latter into one isolated palace or violate the master five layers.
Select the visual-style route outside the canon route. Use
```
华彩通透仙侠
```
by default. Use
```
东方苍穹巨构
```
for monumental architectural fragments, threshold views, divine structures that press against or replace the sky, restrained high-key color, solitary scale markers, and low-angle side-backlight. The scene route and style route are independent additions beneath the master rules.
Select the camera intent. Separate emotional presence from geographic coverage. Default
```
神域聚居地 + 东方苍穹巨构
```
to
```
繁华压迫
```
: a distant low or eye-level camera looking slightly upward, with telephoto compression preferred when dense palace layers must feel heavy. Use a high-angle overview only when explicitly requested.
Choose one spatial skeleton. Use one-point perspective, centered symmetry, diagonal depth, S-curve flow, framed vista, vertical stacking, or another single dominant geometry. Avoid competing perspective systems.
Set the focal hierarchy. For
```
单体仙境
```
, choose one readable landmark. For
```
神域聚居地
```
, choose one principal palace district or divine axis, then organize many subordinate districts beneath it. For
```
东方苍穹巨构
```
, define one dominant spatial event rather than requiring one fully visible object.
Set measurable scale. Start from the four-or-more visible proofs and human comparisons in the master rules, then apply the selected intensity and compatible additions from references/composition-color-light.md. Routes may intensify scale but may not replace measurable evidence with adjectives or density.
Design a chroma budget. Apply the selected saturation strategy and then the selected style route. Keep explicit high-saturation requests, but let
```
东方苍穹巨构
```
preserve a large luminous neutral base and confine the strongest warm color to motivated accents.
Design light. Establish one motivated key-light direction, warm/cool separation, readable shadow color, controlled backlight, and at most one restrained flare or starburst source.
Add material evidence. Describe reflection, translucency, grain, patina, carving depth, edge gilding, moisture, or wear selectively. Prevent plastic-looking jade, metal, glass, and lacquer.
Build depth. Map the scene onto all five master layers. Preserve 40–60% content-bearing air through openings, courts, water, cloud gaps, void, and distance; use high-visibility atmospheric perspective so clouds reveal rather than swallow architecture.
Compose the final prompt. Use the selected route's prompt order. Keep ratio, camera height, viewing distance, tilt, and focal length explicit.
Audit. Run the global inheritance audit first, then the selected route's complete checklist. Do not output until both pass.

锁定场景承诺：总结不可协商的主体、情绪、色调与比例，不丢失明确细节。
应用继承的核心规则：在应用任何路线前，先设定可衡量的巨型尺度依据、完整五层空间、40–60%带内容的呼吸感空间、宏伟东方建筑、无栏杆边缘，以及风格统一的仙人和服饰。后续路线可扩展但不得删除这些设定。
处理独立原典路线：当明确选择
```
仙界大境原典
```
时，保留核心规则，跳过常规场景路线、视觉风格、镜头意图、密度与饱和度推断。应用原典路线更严格的提示词排序，通过两项审核后再输出响应。
否则选择场景路线：单个本地化地点或地标采用
```
单体仙境
```
；有人居住的神界（需保留宫殿区、桥梁、露台、云端基础设施与远处城市层级可见性）采用
```
神域聚居地
```
。不得将后者简化为孤立宫殿，或违反核心五层空间规则。
在原典路线外选择视觉风格路线：默认采用
```
华彩通透仙侠
```
；巨型建筑碎片、门槛视角、贴近或替代天空的神级结构、克制的高亮度色彩、孤立尺度标识、低角度侧逆光场景采用
```
东方苍穹巨构
```
。场景路线与风格路线是核心规则下的独立补充项。
选择镜头意图：区分情感代入与地理覆盖范围。
```
神域聚居地 + 东方苍穹巨构
```
默认采用
```
繁华压迫
```
：远距离低机位或平视机位，略微向上拍摄，若需让密集宫殿层显得厚重，优先采用长焦压缩；桥梁、楼梯、大门与底部展示采用远距离低机位24–35mm镜头；门槛构图采用35–50mm镜头。仅当明确请求时才采用高角度全景。
选择一种空间骨架：采用单点透视、中心对称、对角线纵深、S形流动、框景、垂直堆叠或其他单一主导几何结构，避免冲突的透视系统。
设定焦点层级：
```
单体仙境
```
选择一个清晰可辨的地标；
```
神域聚居地
```
选择一个核心宫殿区或神轴，然后将众多附属区域归属于其下；
```
东方苍穹巨构
```
定义一个主导空间事件，而非要求展示完整物体。
设定可衡量的尺度：从核心规则中的四个及以上可见依据与人物对比入手，然后应用选定的强度与references/composition-color-light.md中的适配补充内容。路线可强化尺度，但不得用形容词或密度替代可衡量的依据。
设计色彩预算：先应用选定的饱和策略，再应用选定的风格路线。保留明确的高饱和请求，但
```
东方苍穹巨构
```
需保留大量明亮中性基底，将最强暖色限制在合理的点缀区域。
设计光线：设定一个合理的主光源方向、冷暖分离、清晰的阴影色彩、可控的逆光，以及至多一个克制的眩光或星芒光源。
添加材质细节：选择性描述反射、半透明、纹理、包浆、雕刻深度、边缘鎏金、湿度或磨损痕迹，避免塑料质感的玉、金属、玻璃与漆器。
构建纵深：将场景映射到所有核心五层空间。通过开口、庭院、水面、云隙、空白与远景保留40–60%带内容的呼吸感空间；采用高可见度空气透视，让云层展现而非遮挡建筑。
撰写最终提示词：采用选定路线的提示词排序，明确标注比例、相机高度、拍摄距离、倾斜角度与焦距。
审核：先进行全局继承规则审核，再进行选定路线的完整清单审核，两项均通过后方可输出。

Independent canon route

独立原典路线

Treat
```
仙界大境原典
```
as a fourth user-facing creative route, not as a private source of the total rules. It inherits the same master scale, space, architecture, railing, figure, and costume language as every other direction.
Never infer nine images, a collage, a sequence, or
```
9:16
```
from this route. Generate one prompt per requested image and support every listed ratio.
Do not import dense-city targets, megastructure-pressure framing, hard side-backlight, high-saturation defaults, or decorative modules from other routes. Railings are already prohibited by the master rules.
Preserve the master giant-scale relationships, five layers, 40–60% content-bearing breathing air, railing-free architecture, back-facing immortals, costume grammar, and physical realism; add the route's stricter low cloud sea, limited far anchors, one primary motion, and warm soft high-altitude diffusion.
If the user asks for several canon prompts, vary composition, main structure, far anchor, ground, figure distance, and motion while keeping every hard invariant unchanged.

将
```
仙界大境原典
```
视为面向用户的第四种创作路线，而非核心规则的私有来源。它与其他所有路线一样，继承相同的核心尺度、空间、建筑、无栏杆、人物与服饰语言。
绝不从该路线推断生成九图、拼贴画、序列或
```
9:16
```
比例，每个请求图像对应一个提示词，支持所有列出的比例。
不得从其他路线导入密集城市目标、巨构压迫式构图、硬侧逆光、高饱和默认值或装饰模块。核心规则已禁止栏杆。
保留核心巨型尺度关系、五层空间、40–60%带内容的呼吸感空间、无栏杆建筑、背影仙人、服饰语法与物理写实性；新增路线更严格的低云海、有限远景锚点、单一主动态与暖柔高空漫射光要求。
若用户请求多个原典提示词，需在保持所有硬性不变项的前提下，更改构图、主体结构、远景锚点、地面、人物距离与动态。

Scene-route selection

场景路线选择

Select
```
单体仙境
```
for a corridor, gate, pavilion, ritual terrace, isolated floating island, solitary mountain, sacred tree, moon ring, or another scene organized around one localized experience. Preserve one ruling anchor while mapping it onto the master five layers and 40–60% content-bearing air.
Select
```
神域聚居地
```
for
```
天宫
```
,
```
仙城
```
,
```
神都
```
,
```
九重天
```
,
```
云上帝都
```
,
```
神域
```
,
```
神仙居所
```
,
```
宫阙群
```
, a dense inhabited floating civilization, or a request for a city-scale divine world. Load and follow references/celestial-realm-route.md.
Treat explicit user layout, lens, depth of field, density, placement, palette, and lighting as hard locks. Route selection must not replace them with a familiar corridor, moon gate, centered palace, or other house composition.
Interpret traditional Chinese architecture as a recognizable design language, not a reconstruction limit. Permit physically impossible floating continents, cloud-borne infrastructure, vertical divine cities, celestial waterfalls, and nested heavens while keeping perspective, light, material response, and internal spatial logic coherent.

廊道、大门、亭台、祭台、孤立浮岛、孤山、神树、月环或其他围绕单个本地化体验构建的场景，选择
```
单体仙境
```
。保留一个主导锚点，同时将其映射到核心五层空间与40–60%带内容的呼吸感空间。
```
天宫
```
、
```
仙城
```
、
```
神都
```
、
```
九重天
```
、
```
云上帝都
```
、
```
神域
```
、
```
神仙居所
```
、
```
宫阙群
```
、密集的浮空文明或城市级神界请求，选择
```
神域聚居地
```
，并加载遵循references/celestial-realm-route.md。
将用户明确指定的布局、镜头、景深、密度、位置、色调与光线视为硬性锁定项。路线选择不得用熟悉的廊道、月门、中心宫殿或其他固定构图替代这些设定。
将中国传统建筑视为可识别的设计语言，而非复刻限制。允许物理上不可能的浮空大陆、云端基础设施、垂直神城、天瀑与嵌套天界，但需保持透视、光线、材质反馈与内部空间逻辑连贯。

Visual-style route selection

视觉风格路线选择

Select
```
华彩通透仙侠
```
by default for richly colored materials, intricate ornament, clear architectural presentation, luminous clouds, and elegant cinematic spectacle.
Select
```
东方苍穹巨构
```
when the user requests or references
```
苍穹巨构
```
,
```
白玉京巨构
```
,
```
巨物压迫
```
,
```
建筑切片
```
,
```
巨构压顶
```
,
```
门槛窥视
```
,
```
不可抵达的天宫
```
,
```
孤寂神圣
```
, restrained blue-gray-and-gold light, or a visual reference dominated by those properties. Load and follow references/eastern-sky-megastructure-style.md.

Keep both route axes explicit. Valid combinations include

单体仙境 + 华彩通透仙侠

单体仙境 + 东方苍穹巨构

神域聚居地 + 华彩通透仙侠

, and

神域聚居地 + 东方苍穹巨构

Treat a dominant spatial event as one interaction such as a city beneath an overhanging gate, a bridge vanishing into an unreachable palace, a waterfall rupturing a sky-city, or a distant realm revealed through a portal. Multiple elements may form the event without becoming competing focal points.

默认采用
```
华彩通透仙侠
```
，适用于色彩丰富的材质、复杂纹饰、清晰的建筑呈现、明亮云层与优雅的电影级奇观。
当用户请求或提及
```
苍穹巨构
```
、
```
白玉京巨构
```
、
```
巨物压迫
```
、
```
建筑切片
```
、
```
巨构压顶
```
、
```
门槛窥视
```
、
```
不可抵达的天宫
```
、
```
孤寂神圣
```
、克制的蓝灰金光线，或视觉参考以这些特征为主时，选择
```
东方苍穹巨构
```
，并加载遵循references/eastern-sky-megastructure-style.md。

明确保留两条路线轴，有效组合包括

单体仙境 + 华彩通透仙侠

、

单体仙境 + 东方苍穹巨构

、

神域聚居地 + 华彩通透仙侠

、

神域聚居地 + 东方苍穹巨构

。

主导空间事件指一种互动场景，如悬门下方的城市、消失于遥不可及宫殿的桥梁、冲破天空之城的瀑布，或通过门户展现的远方领域。多个元素可构成事件，但不得成为相互竞争的焦点。

Camera-intent routing

镜头意图路由

Preserve an explicit camera direction as a hard lock.
Use
```
繁华压迫
```
when the goal is awe, dominance, divine prosperity, architectural weight, or the feeling of standing inside the realm. Keep the camera low or at human eye level, look level or slightly upward, keep the horizon low, and let the principal structure occupy the middle and upper frame. Prefer a distant 70–100mm view to compress dense palace layers; use a distant low 24–35mm view for bridges, stairs, gates, and underside reveals; use 35–50mm for threshold framing.
Use
```
沉浸探索
```
for corridors, terraces, gardens, bridges, interiors, and journeys through the world. Keep the camera near human height and choose focal length from the spatial path.
Use
```
城市总览
```
only when the user explicitly requests
```
俯拍
```
,
```
鸟瞰
```
,
```
航拍
```
,
```
城市总览
```
,
```
地图感
```
,
```
城区分布
```
,
```
浮岛关系
```
, or equivalent geographic information. This is the only route that may default to an elevated camera.
Treat
```
远距离
```
and
```
高机位
```
as independent. A camera can remain low while standing far from the city, and that is the preferred setup for monumental pressure.
Do not infer an elevated view from
```
大全景
```
,
```
云海
```
,
```
仙城
```
,
```
浮岛
```
,
```
世界观
```
, or a wide aspect ratio alone.

将明确的相机方向视为硬性锁定项。
当目标是震撼感、统治感、神界繁华、建筑厚重感或置身于天界的感受时，采用
```
繁华压迫
```
。保持相机低机位或人眼高度，平视或略微向上拍摄，压低地平线，让主体结构占据中上部画面。密集宫殿层优先采用70–100mm远距离镜头压缩；桥梁、楼梯、大门与底部展示采用24–35mm远距离低机位镜头；门槛构图采用35–50mm镜头。
廊道、露台、花园、桥梁、室内与穿越世界的旅程采用
```
沉浸探索
```
。保持相机接近人眼高度，根据空间路径选择焦距。
仅当用户明确请求
```
俯拍
```
、
```
鸟瞰
```
、
```
航拍
```
、
```
城市总览
```
、
```
地图感
```
、
```
城区分布
```
、
```
浮岛关系
```
或类似地理信息时，采用
```
城市总览
```
。这是唯一可默认采用高机位的路线。
将
```
远距离
```
与
```
高机位
```
视为独立项。相机可保持低机位同时远离城市，这是营造巨型压迫感的首选设置。
不得仅从
```
大全景
```
、
```
云海
```
、
```
仙城
```
、
```
浮岛
```
、
```
世界观
```
或宽屏比例推断高角度视角。

Aspect-ratio rules

比例规则

Put the selected ratio in the first sentence of every copy-ready prompt.
Translate the ratio into compositional behavior using references/aspect-ratios.md; do not append it as a meaningless tag.
Keep exact pixel dimensions separate from aspect ratio when the user provides both.
If the user names a generation platform, add its native ratio syntax only when supported and known. Do not invent platform flags.
Preserve the same ratio across prompt iterations unless the user changes it.

在每个可直接使用的提示词首句标注选定比例。
使用references/aspect-ratios.md将比例转化为构图行为，不得将其作为无意义标签附加。
当用户同时提供精确像素尺寸与比例时，将两者分开记录。
若用户指定生成平台，仅在支持且已知的情况下添加其原生比例语法，不得自创平台标识。
除非用户更改，否则在提示词迭代过程中保留相同比例。

Scale and saturation routing

尺度与饱和度路由

Use
```
辽阔
```
for landscape-led scenes whose giant structure is quieter and whose air dominates, while retaining all five layers, 40–60% content-bearing air, and four measurable scale proofs.
Use
```
史诗级
```
by default. Strengthen the ruling subject and discontinuous comparisons without dropping the master baseline.
Use
```
超宏大
```
only when the frame passes both the master audit and the compatible macro additions in references/composition-color-light.md: tiny adults, an off-frame or cross-cloud giant form, a readable extreme distance, and stronger scale discontinuity.
For
```
神域聚居地
```
, map the same master rules onto a ruling district, connected subordinate districts, human-bearing circulation, and an extreme-distance settlement. Density never replaces the fixed layers, measurable comparisons, or 40–60% content-bearing air.
Use
```
自然鲜明
```
for realistic medium chroma with one restrained accent.
Use
```
选择性高饱和
```
by default. Concentrate high chroma in a small accent area while keeping large light surfaces and distance calmer.
Use
```
华丽高饱和
```
when explicitly requested. Preserve hierarchy by limiting the number of equally intense hues and retaining neutral anchors.
Do not change one routing value because of another. A scene may be
```
超宏大 + 自然鲜明
```
or
```
辽阔 + 华丽高饱和
```
.

```
辽阔
```
适用于景观主导的场景，其巨型结构相对低调，空间占主导，同时保留所有五层空间、40–60%带内容的呼吸感空间与四个可衡量的尺度依据。
默认采用
```
史诗级
```
，强化主导主体与非连续对比，同时不低于核心基准线。
仅当画面通过核心规则审核与references/composition-color-light.md中的适配宏观补充内容（微小人物、画外或跨云巨型形态、清晰的极端远景、更强的尺度断层）时，采用
```
超宏大
```
。
对于
```
神域聚居地
```
，将相同核心规则映射到主导区域、相连附属区域、有人活动的流通空间与极端远景聚居地。密度不得替代固定空间层、可衡量对比或40–60%带内容的呼吸感空间。
```
自然鲜明
```
适用于写实中等色彩饱和度，搭配一个克制的点缀色。
默认采用
```
选择性高饱和
```
，将高色彩饱和度集中在小范围点缀区域，同时保持大面积亮面与远景色调柔和。
仅当明确请求时采用
```
华丽高饱和
```
，通过限制同等强度色调的数量并保留中性锚点来维持层级。
不得因一个路由值更改另一个，场景可以是
```
超宏大 + 自然鲜明
```
或
```
辽阔 + 华丽高饱和
```
。

Visual constraints

视觉约束

Preserve explicit high-saturation or bright-color requirements, but translate them through the selected saturation strategy.

Prefer

高明度、选择性高纯度、色彩分离清楚、阴影透明、亮而不荧光

over a bare

高饱和

instruction.

Do not apply a global orange-gold cast. Keep white jade, cloud highlights, skin, and pale stone materially distinct; confine warm spill to motivated lit surfaces.
Do not claim an image is monumental without visible scale evidence. Apply the selected scale gate before finalizing.
Do not default every scene to white jade corridors, vermilion columns, circular moon gates, three women, or blue sky. Those are optional modules, not the Skill's identity.
Use no railing, balustrade, fence, parapet, guardrail, or decorative edge barrier unless the user explicitly requests one.
In
```
单体仙境
```
, do not combine several dominant landmarks. In
```
神域聚居地
```
, allow many buildings and wonders but subordinate them to one city hierarchy and one coherent Eastern architectural language.
Do not let haze create a gray veil, flatten color, or erase distant silhouettes.
Do not overuse shallow depth of field in environment-wide shots. Default
```
神域聚居地
```
establishing shots to deep focus with readable foreground, midground, background, and extreme-distance architecture unless the user explicitly requests otherwise.
In
```
东方苍穹巨构
```
, do not shrink the main structure to show it completely. Preserve deliberate cropping, occlusion, threshold framing, and at least one discontinuous scale jump.
Do not use high-angle, bird's-eye, aerial, or top-down views for
```
繁华压迫
```
. Do not sacrifice vertical faces, overhanging mass, or the viewer's subordinate position merely to show the whole city.
Keep lens flare subtle and optically motivated.
Avoid copyrighted artist-name imitation. Describe visual properties directly.
Default people to fictional adults at master-rule scale, rear-facing or rear-oblique, with calm actions, structured immortal clothing, and one coherent wind direction. Preserve reference identity only for user-owned or authorized images.

保留明确的高饱和或亮色要求，但需通过选定的饱和策略转化。

优先采用

高明度、选择性高纯度、色彩分离清楚、阴影透明、亮而不荧光

，而非单纯的

高饱和

指令。

不得应用全局橙金色调，保持白玉、云层高光、皮肤与浅色石材的材质区分；将暖色漫射限制在合理的受光面。
若无可见尺度依据，不得声称图像具有巨型感，在最终确定前应用选定的尺度门槛。
不得将所有场景默认设置为白玉廊道、朱红立柱、圆形月门、三位女性或蓝天，这些是可选模块，而非本Skill的固有特征。
除非用户明确请求，否则不得使用栏杆、扶手、围栏、护墙、护栏或装饰性边缘屏障。
```
单体仙境
```
不得组合多个主导地标；
```
神域聚居地
```
允许存在众多建筑与奇观，但需归属于一套城市层级与统一的东方建筑语言。
不得让雾气形成灰色面纱、 flatten色彩或消除远景轮廓。
环境全景镜头不得过度使用浅景深，
```
神域聚居地
```
开场镜头默认采用深景深，清晰呈现前景、中景、背景与极端远景建筑，除非用户明确请求其他设置。
```
东方苍穹巨构
```
不得缩小主体结构以完整展示，需保留刻意裁剪、遮挡、门槛构图与至少一处尺度断层。
```
繁华压迫
```
不得使用高角度、鸟瞰、航拍或俯视视角，不得为展示整个城市而牺牲垂直面、悬挑体量或观者的从属位置。
镜头眩光需柔和且符合光学逻辑。
避免模仿受版权保护的艺术家风格，直接描述视觉特征。
默认人物为符合核心规则尺度的虚构成年人，背向或侧背向，动作平静，穿着结构规整的仙服，风向统一。仅保留用户自有或授权图像的参考特征。

Standard output

标准输出

Unless the user asks for another mode, return:

```
参数锁定
```
: compact field list, including
```
共享总则继承
```
, the exact or supplemented ratio, independent canon route when selected, otherwise scene route and visual-style route, camera intent, camera height and lens, scale intensity, and saturation strategy.
```
视觉导演方案
```
: short explanation of composition, palette, light, scale, and atmosphere.
```
完整提示词
```
: one copy-ready fenced
```
text
```
block.
```
负面约束
```
: a separate fenced
```
text
```
block.
```
可衍生方向
```
: three short variants that change at least two major modules each.

For

只要提示词

or concise mode, return only the complete prompt and negative constraints. For variants mode, give each variant its own complete prompt when requested; otherwise provide short direction cards.

除非用户要求其他模式，否则返回：

```
参数锁定
```
：紧凑字段列表，包括
```
共享总则继承
```
、精确或补充的比例、选定的独立原典路线（若有），否则为场景路线与视觉风格路线、镜头意图、机位与镜头、尺度强度、饱和策略。
```
视觉导演方案
```
：关于构图、色调、光线、尺度与氛围的简短说明。
```
完整提示词
```
：一个可直接使用的 fenced
```
text
```
块。
```
负面约束
```
：一个单独的 fenced
```
text
```
块。
```
可衍生方向
```
：三个简短变体，每个至少更改两个核心模块。

若用户要求

只要提示词

或简洁模式，仅返回完整提示词与负面约束。若为变体模式，当用户请求时为每个变体提供单独的完整提示词；否则提供简短方向卡片。

Prompt construction requirements

提示词构建要求

The complete prompt must:

Begin with

<ratio> 画幅，<framing/camera>，<dominant composition>

Name
```
仙界大境原典
```
near the beginning when that independent route is selected, then satisfy its route-specific assembly and audit instead of ordinary style defaults.
Express the selected scale intensity through visible scale relationships rather than an adjective alone.
Assign all five master layers and state how 40–60% content-bearing breathing space remains readable.
State where the landmark is located for
```
单体仙境
```
, or how the principal district governs subordinate city layers for
```
神域聚居地
```
.
For
```
东方苍穹巨构
```
, state the dominant spatial event, what hides or crops the megastructure, and how the figure, threshold, or distant city proves its impossible size.
State camera height, viewing distance, tilt, and focal length separately. Never use
```
大全景
```
or
```
建立镜头
```
as a substitute for camera placement.
Explain how scale decreases into depth when strict perspective matters.
Include at least four measurable scale proofs and keep all edges railing-free unless explicitly overridden.
Name the palette, high-chroma accent, neutral anchors, and shadow color separately.
Use a single dominant light source and describe its direction.
When people are present, state figure percentage, rear-facing angle, calm action, immortal-costume structure, and one wind direction.
Place style and rendering language after the concrete scene description.
End with critical invariants and exclusions.

Keep positive and negative instructions separate. Favor concrete visible relationships over adjective stacking.

完整提示词必须：

以

<ratio> 画幅，<framing/camera>，<dominant composition>

开头。

当选定独立原典路线时，在开头附近标注
```
仙界大境原典
```
，然后满足该路线特定的组装与审核要求，而非常规风格默认值。
通过可见尺度关系表达选定的尺度强度，而非仅使用形容词。
分配所有核心五层空间，并说明40–60%带内容的呼吸感空间如何保持清晰可见。
说明
```
单体仙境
```
的地标位置，或
```
神域聚居地
```
的核心区域如何管控附属城市层级。
对于
```
东方苍穹巨构
```
，说明主导空间事件、遮挡或裁剪巨构的元素，以及人物、门槛或远方城市如何证明其超乎寻常的尺度。
分别记录相机高度、拍摄距离、倾斜角度与焦距，不得用
```
大全景
```
或
```
建立镜头
```
替代机位设置。
当严格透视重要时，说明尺度如何随纵深递减。
包含至少四个可衡量的尺度依据，除非明确覆盖，否则所有边缘无栏杆。
分别标注色调、高饱和度点缀色、中性锚点与阴影色彩。
使用单一主光源并描述其方向。
当有人物时，说明人物比例、背向角度、平静动作、仙服结构与统一风向。
将风格与渲染语言放在具体场景描述之后。
以关键不变项与排除项结尾。

将正面与负面指令分开，优先使用具体可见关系而非堆砌形容词。