layer-game-assets

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Layer Game Assets

Layer 游戏资产

Overview

概述

Game art differs from illustration in one way that governs everything: it has to sit in a system. An icon shares a set, a character shares a cast, a background shares a parallax stack. A beautiful asset that does not match its neighbours is a defect.

Generation mechanics are in the

layer

skill and prompt craft is in

layer-image

. This skill covers what those leave out: the constraints specific to art that ships inside a game. Pixel art has its own rules (

layer-pixel-art

), as do tiling surfaces (

layer-textures

If a sibling skill named here is missing from your available skills, ask the user to install it (

npx skills add layerai/skills --skill <name>

); unattended, proceed from tool schemas and flag the gap.

游戏美术与插画有一处差异，并决定了一切：它必须融入一套系统。图标要融入图标集，角色要融入角色阵容，背景要融入视差图层堆栈。不匹配相邻素材的精美资产就是缺陷。

layer

skill 负责生成机制，

layer-image

负责提示词打磨。本 skill 补充两者未涵盖的内容：游戏内发布美术特有的约束。像素美术有独立的规则（

layer-pixel-art

），可平铺表面亦然（

layer-textures

）。

如果这里提到的同级 skill 不在可用 skills 中，请用户安装（

npx skills add layerai/skills --skill <name>

）；在无人值守模式下，则根据 tool schema 继续，并指出这一缺失。

Characters and casts

角色与角色阵容

A character prompt must name the role before the look. An enemy grunt, a shopkeeper, and a raid boss are read differently by players, and silhouette carries that read before any detail does. State silhouette explicitly: bulk, stance, and the one shape that identifies them at a glance.

Then pin the production terms a prompt cannot leave to chance: full body or portrait, view (front, three-quarter, side), background treatment (isolated on flat colour if it will be composited), and whether the pose is neutral for reference or dynamic for key art.

A cast is the hard part. Writing the same adjectives twice does not produce the same character, and seeds do not carry across prompts. Two things work:

Anchor image. Approve one hero image, then drive every sibling with it attached as
```
reference_image
```
, filtering for
```
image_editing
```
, which is the capability for art that must reproduce a provided subject.
```
style_reference
```
transfers a look only and will not keep the character, which is the usual reason an anchored cast still drifts.
Trained style. For a cast that will grow over weeks, train a reference set on approved art and apply it to every run. See
```
layer-reference-sets
```
.

The first is right for a handful of assets this week. The second is right for a live game.

角色提示词必须先说明角色定位，再描述外观。敌人杂兵、商店老板和袭击首领在玩家眼中有不同解读，而轮廓会先于任何细节传达这种差异。明确说明轮廓：体量、站姿，以及让人一眼识别的关键形状。

随后，需要明确提示词中不能碰运气的制作规格：全身或肖像、视角（正面、四分之三侧面或正侧面）、背景处理方式（如果素材需要合成，则使用纯色背景进行隔离），以及姿势是用于参考图的中性姿势，还是用于主视觉的动态姿势。

角色阵容是最难的部分。两次使用相同的形容词无法生成同一个角色，seed 也不会在不同提示词之间延续。有两种方法有效：

锚定图像（Anchor image）。 先批准一张主角图，然后将其作为
```
reference_image
```
附加到每个同系列素材的生成中，并筛选具备
```
image_editing
```
能力的模型；这项能力专门用于复现已提供主体的美术。
```
style_reference
```
只会迁移风格，无法保持角色本身，而这通常正是使用锚定方法后角色阵容仍发生漂移的原因。
训练风格（Trained style）。 对于将在数周内持续扩充的角色阵容，基于已批准的美术作品训练一个参考集，并将其应用到每次生成中。参见
```
layer-reference-sets
```
。

第一种方法适合本周处理少量素材，第二种方法适合长期运营的游戏。

Rig-ready characters

适用于骨骼 rig 的角色

A character destined for Spine or a skeletal rig needs its parts separable. Generate the character in a neutral, limbs-apart pose so no part occludes another, on a transparent or flat background, then separate the parts with

layer-image-editing

using

split_into_layers

Occlusion is the thing to get right up front: a crossed arm or a cloak covering the torso means the hidden geometry does not exist in any layer, and no amount of splitting recovers it. Say "arms clear of the body, legs apart, cape swept behind and clear of the torso" in the prompt.

用于 Spine 或骨骼 rig 的角色需要各部件可分离。以中性、四肢分开的姿势生成角色，确保任何部件都不会遮挡其他部件，并使用透明或纯色背景；随后使用

layer-image-editing

的

split_into_layers

拆分部件。

遮挡是必须在一开始就处理正确的问题：交叉的双臂或遮住躯干的斗篷意味着隐藏的几何结构不会存在于任何图层中，再怎么拆分也无法恢复。可在提示词中写："双臂离开身体，双腿分开，斗篷向后甩开且不遮挡躯干"。

Environments and parallax

环境与视差

An environment prompt needs camera height and depth layers stated, because these are what make a background sit behind gameplay rather than compete with it. Name the tier the art belongs to: foreground silhouette, midground playable, far background.

For a parallax stack, generate each layer as a separate run against one shared style anchor, and say in each prompt what that layer contains and that the rest is empty: "far background only, distant mountains and sky, nothing in the midground or foreground, isolated on transparent". Generating the whole scene and splitting it afterwards gives layers that were never designed to move independently.

Keep contrast and saturation deliberately low on gameplay backgrounds. The most common note on generated game backgrounds is that they are too busy to read characters against.

环境提示词需要说明摄像机高度和深度图层，因为正是这些因素让背景处于游戏玩法之后，而不是与玩法争夺注意力。注明美术所属的层级：前景剪影、中景可玩区域或远景背景。

对于视差堆栈，基于同一个共享风格锚点分别生成每一层，并在每个提示词中说明该层包含什么、其他部分为空："仅远景背景，包含远山与天空，中景和前景均为空，透明背景上孤立呈现"。先生成完整场景再拆分为图层，会得到从未为独立运动而设计的图层。

有意降低游戏背景的对比度和饱和度。生成的游戏背景最常见的问题反馈是：画面过于繁杂，无法凸显角色。

UI and icons

UI 与图标

State the pixel size in the prompt, and design for the smallest one. An icon set delivered at 512px that turns to mush at 64px has failed, and the fix is fewer elements and heavier strokes, not a better upscale.

Hold a set together by naming the shared grammar in every prompt: stroke weight, corner radius, perspective (flat, isometric, or slight three-quarter), palette, and whether icons sit on a plate or float free. Generate on transparent backgrounds so the engine can place them.

Extending an existing set is an editing job, not a generation job: attach two or three approved icons as style references so the new one inherits the grammar rather than reinventing it.

在提示词中注明像素尺寸，并按最小尺寸进行设计。交付时为 512px、到 64px 时却模糊成一团的图标集属于失败；解决办法是减少元素并加粗描边，而不是进行更好的放大处理。

通过在每个提示词中明确共享的视觉语法来保持整套图标一致：描边粗细、圆角、透视（平面、等距或轻微四分之三视角）、调色板，以及图标是置于底板上还是独立浮动。在透明背景上生成，以便引擎进行放置。

扩展现有图标集属于编辑任务，而不是生成任务：附加两到三个已批准的图标作为 style reference，让新图标继承既有的视觉语法，而不是重新创造一套规范。

Worked example

实例

"Six ability icons for our fantasy RPG, matching the two we already approved."

The two approved icons already have
```
file_id
```
values from their original runs.

list_base_models

with

filter.use_case: "text_to_image"

and

filter.capabilities: {image_editing: true, square: true}

(an object of booleans, not a list of names). Take the first result.

One run per icon, six runs. Each attaches both approved icons as
```
reference_image
```
, and each prompt names the shared grammar plus the one new subject: "Ability icon, 64px target, flat three-quarter view, heavy uniform outline, two-tone fill, muted gold and slate palette, isolated on transparent. Subject: a frost sigil."
Estimate the first, confirm the look, then run the remaining five.
Review them together at 64px, not individually at full size.

"为我们的奇幻 RPG 制作六个能力图标，与已经批准的两个图标保持一致。"

两个已批准的图标已经拥有来自最初生成的
```
file_id
```
值。

调用

list_base_models

，设置

filter.use_case: "text_to_image"

和

filter.capabilities: {image_editing: true, square: true}

（这是布尔值对象，而不是名称列表）。取第一个结果。

每个图标单独运行一次，共六次。每次都附加两个已批准的图标作为
```
reference_image
```
，并在每个提示词中注明共享的视觉语法和唯一的新主体："能力图标，目标尺寸 64px，平面四分之三视角，粗重且统一的轮廓，双色填充，低饱和度的金色与石板色调色板，透明背景上孤立呈现。主体：霜冻徽记。"
先运行第一个，确认效果后，再运行其余五个。
在 64px 下统一审阅，而不是在全尺寸下逐个审阅。

Common mistakes

常见错误

Judging an icon at full resolution instead of at its target size.
Rewriting the same description and expecting the same character.
Generating a whole scene, then splitting it into parallax layers afterwards.
Generating a rig-ready character in a pose where limbs occlude the torso.
Leaving background treatment unstated on an asset that will be composited.
High-contrast, busy gameplay backgrounds that characters cannot read against.
Adding to an approved icon set without attaching the approved icons as references.

在全尺寸下判断图标，而不是在目标尺寸下判断。
重复改写相同的描述，却期望得到同一个角色。
先生成完整场景，然后再拆分为视差图层。
在四肢会遮挡躯干的姿势下生成适用于骨骼 rig 的角色。
对于需要合成的素材，未说明背景处理方式。
游戏背景对比度过高且过于繁杂，导致角色无法从中凸显。
扩展已批准的图标集时，未将已批准的图标作为参考图像附加进去。