F
fabriciof807/pixel-art-skills
Claude skills: 3D model (Blender) to isometric pixel art sprite pipeline
更新 2026-11-16
在 GitHub 查看 ↗链接已于 2026-11-17 核查 · HTTP 200 · 打开的是厂商自己的官网
装完之后怎么算装好了
- 装完该看到什么 — 命令正常结束时退出码为 0,终端里没有 "command not found" 或 "permission denied"。
- 一行自检 — 按该仓库 README 里给出的最小用法跑一次;能按 README 的描述输出就是通了。
- 答不出来怎么办 — 这个项目没有写排错说明,去仓库的 Issues 搜 install。
安装命令取自该仓库;以上判据只描述可观察到的结果,不替代项目自己的文档。
- ★ 数取值
- 0 ★ · fabriciof807/pixel-art-skills · 取自 GitHub 搜索 API · (2026-11-19 20:00:48 核对)
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型:开源项目
- 收录与活跃
- 收录于 2026-11-17 · 最近见到 2026-11-20 · 共见到 4 次 · 被 2 个来源列出
- 评分构成
- 综合分 203 · 2 个来源 · repo
- 安装
- 安装命令见上方「安装与配置」。
常见问题(答案都来自我们的记录)
它适合谁 / 用来做什么?
我们把它归在这些标签下: 开源项目。标签来自条目自身,不是我们强加的分类。
谁维护它、还在更新吗?
仓库
fabriciof807/pixel-art-skills;我们收录于 2026-11-17;最近见到 2026-11-20。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。README 原文(英文,未翻译) · 在 GitHub 读完整 README →
# pixel-art-skills
以下摘录自该仓库的 README(片段)
Six skills for turning a 3D model (Blender) into isometric pixel art sprites:
8 directions, several animations, one consistent style across a whole series.
Based on the pipeline from Mekai Games, "How to Transform a 3D Model Into Pixel Art"
(https://www.youtube.com/watch?v=Y4uTF3aBTKo).
## Install
With skills, on any machine:
bash
# all six skills, globally (every project)
npx skills add fabriciof807/pixel-art-skills -g -y
# only one skill
npx skills add fabriciof807/pixel-art-skills -g -y --skill pixel-art-cleanup
# list what the repo contains
npx skills add fabriciof807/pixel-art-skills --list
Drop -g to install into the current project only. Or copy the skills/ folders into
~/.claude/skills/ by hand.
## Skills
### pixel-art-3d-pipeline: overview
Entry point. Explains the whole pipeline, the order of the other five skills, and the
project parameters that are decided once and written into the project's art document:
sprite height (64 px in the video), directions (8), animations (idle, walk, attack, hit,
death), palette (26 colors in the video), and light direction (upper left).
Its core principle: the pipeline produces frames; it does not judge them. Every
character is still reviewed frame by frame, and fixes go back to the model.
### pixel-art-modeling: step 1, per character
Builds the character from simple shapes (boxes, cylinders, spheres). At 64 px tall,
detail becomes noise, so the silhouette is what matters: one strong, instantly
readable shape per character, plus the family traits shared by the whole series.
Keeps parts as separate objects and materials, because the part-ID pass and the contour
lines depend on them.
### pixel-art-camera-lighting: step 2, once per project
Sets up the camera and light that every character shares:
Orthographic camera, no perspective, tilted down 30°. Diagonals then step 2 px across
per 1 px down, the classic 2:1 isometric grid. The skill notes that the exact 2:1 slope
is about 26.57°; if diagonals still look off at sprite size, test that value.
Toon shader with three tones per material (shadow, mid, highlight). Shadows lean cool,
highlights lean warm.
One sun from the upper left, for every character. No soft light, ambient occlusion,
bloom, or anti-aliasing on the final render.
### pixel-art-render-directions: step 3, per character
Rotates the model, not the camera: eight steps of 45° around the vertical axis, so the
light stays fixed. For each direction it renders every animation frame at the final sprite
size, plus a second hidden pass with a flat unique color per part (the part-ID mask).
The output is a raw render: close, but not pixel art yet. It is not shipped as-is. The
skill suggests a predictable naming scheme ( / / _ .png and
..._id.png) and a script for the loop, since one character is around 400 frames.
### pixel-art-cleanup: step 4, per character
A script applied to every frame with identical rules, in this order:
1. Palette snap: every pixel goes to the nearest color of the project's fixed palette.
2. Remove stray single pixels, which read as dirt.
3. Part contours: thin lines where two parts overlap, using the part-ID mask.
4. One-pixel dark outline around the whole sprite (after the contours, so they are not
8 directions, several animations, one consistent style across a whole series.
Based on the pipeline from Mekai Games, "How to Transform a 3D Model Into Pixel Art"
(https://www.youtube.com/watch?v=Y4uTF3aBTKo).
## Install
With skills, on any machine:
bash
# all six skills, globally (every project)
npx skills add fabriciof807/pixel-art-skills -g -y
# only one skill
npx skills add fabriciof807/pixel-art-skills -g -y --skill pixel-art-cleanup
# list what the repo contains
npx skills add fabriciof807/pixel-art-skills --list
Drop -g to install into the current project only. Or copy the skills/ folders into
~/.claude/skills/ by hand.
## Skills
### pixel-art-3d-pipeline: overview
Entry point. Explains the whole pipeline, the order of the other five skills, and the
project parameters that are decided once and written into the project's art document:
sprite height (64 px in the video), directions (8), animations (idle, walk, attack, hit,
death), palette (26 colors in the video), and light direction (upper left).
Its core principle: the pipeline produces frames; it does not judge them. Every
character is still reviewed frame by frame, and fixes go back to the model.
### pixel-art-modeling: step 1, per character
Builds the character from simple shapes (boxes, cylinders, spheres). At 64 px tall,
detail becomes noise, so the silhouette is what matters: one strong, instantly
readable shape per character, plus the family traits shared by the whole series.
Keeps parts as separate objects and materials, because the part-ID pass and the contour
lines depend on them.
### pixel-art-camera-lighting: step 2, once per project
Sets up the camera and light that every character shares:
Orthographic camera, no perspective, tilted down 30°. Diagonals then step 2 px across
per 1 px down, the classic 2:1 isometric grid. The skill notes that the exact 2:1 slope
is about 26.57°; if diagonals still look off at sprite size, test that value.
Toon shader with three tones per material (shadow, mid, highlight). Shadows lean cool,
highlights lean warm.
One sun from the upper left, for every character. No soft light, ambient occlusion,
bloom, or anti-aliasing on the final render.
### pixel-art-render-directions: step 3, per character
Rotates the model, not the camera: eight steps of 45° around the vertical axis, so the
light stays fixed. For each direction it renders every animation frame at the final sprite
size, plus a second hidden pass with a flat unique color per part (the part-ID mask).
The output is a raw render: close, but not pixel art yet. It is not shipped as-is. The
skill suggests a predictable naming scheme ( / / _ .png and
..._id.png) and a script for the loop, since one character is around 400 frames.
### pixel-art-cleanup: step 4, per character
A script applied to every frame with identical rules, in this order:
1. Palette snap: every pixel goes to the nearest color of the project's fixed palette.
2. Remove stray single pixels, which read as dirt.
3. Part contours: thin lines where two parts overlap, using the part-ID mask.
4. One-pixel dark outline around the whole sprite (after the contours, so they are not
收录徽章(可选)
想把收录标记放进 README 就复制这行,可选,不影响收录。
[](https://anclum.com/zh/s/fabriciof807-pixel-art-skills/)
- 来源
- github · https://github.com/fabriciof807/pixel-art-skills
- 按领域浏览
- 开源项目
- 收录来源
- 2 个独立来源
- 收录于
- 今天
更多条目
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。