agentleFS
dlazy-aidlazy-ai/ecommerce-skills/skills/clothing-detail/skill.md

clothing-detail

Skill71 starsChanged 3 months agoMITView on GitHub

服装工艺细节放大图。服装图 → 面料纹理、走线、织法的微距特写。当用户说「细节图」「特写」「面料放大」「工艺展示」「近景细节」时使用。

The file

Files
skill.md

clothing-detail — 服装图生成细节放大图

一张服装图 → 局部微距特写。详情页里「证明这件衣服做得好」的那几张图。

为什么需要:转化率高的详情页通常有 2~3 张细节图(领口、袖口、面料纹理),但拍微距要专门的镜头和布光。本技能从常规商品图推出这些特写。


生成效果示例

输入:服装图
<img src="../../docs/clothing-detail/garment-flatlay.jpg" width="280">
garment-flatlay.jpg — 军绿麻花针织毛衣平铺图,800×800

实际执行的命令:

dlazy gpt-image-2 \
  --prompt 'Macro detail shot for an e-commerce detail page. Zoom into the shoulder-and-collar area of this olive-green cable-knit sweater and render a photorealistic close-up that fills the frame. Show the ribbed crewneck collar meeting the raglan-style cable panel, individual yarn plies and the twist of the cable braid, the loft of the wool fibres, and soft directional light raking across the surface to reveal depth. Keep the colour and stitch pattern exactly as in the source. Shallow depth of field with the far edge softly out of focus, clean neutral background bokeh, no person, no text, no watermark.' \
  --images docs/clothing-detail/garment-flatlay.jpg \
  --size 1024x1024 --quality high --imageFormat jpeg \
  --save docs/clothing-detail/example-output.jpg

输出

<img src="../../docs/clothing-detail/example-output.jpg" width="320">

example-output.jpg — 1024×1024,60 credits。领口罗纹与麻花panel的交接、每根纱线的捻向、羊毛纤维的绒毛感都被解析出来,侧光让菱形提花的凹凸立体可见,远端落入柔和虚化。


1、能力边界

能力说明
取景部位领口罗纹 / 袖口 / 下摆 / 纽扣 / 拉链 / 口袋 / 刺绣 / 印花 / 织法结构 / 面料纤维
风格控制参考图(照抄某张细节图的机位与光线)或自定义提示词
服装类型帮助模型判断哪些部位值得放大
生成比例1:1(方图细节位)/ 3:4(竖版详情页)

不做:不改颜色、织法与结构;不添加原图没有的工艺(不存在的刺绣、不存在的拉链);不虚构面料成分。


2、输入素材规则

生成前先自检这几条硬性约束:

  • 大小:20KB ~ 15MB
  • 分辨率:大于 400×400
  • 格式:jpg / jpeg / png / webp

输入建议

做法说明
✅ 原图分辨率越高越好微距是在放大原图信息,原图糊 = 细节图编
✅ 目标部位在原图里清晰可见原图里看不清的部位,输出的是模型的想象
✅ 一次只放大一个部位一张图里塞三个特写等于都不清楚
❌ 低分辨率 / 强压缩图会放大出塑料感的假纹理
❌ 目标部位被遮挡挡住的工艺只能靠编

3、取景部位 → prompt 写法

部位取景描述
领口罗纹the ribbed crewneck collar meeting the body panel, showing rib wale spacing and the seam join
袖口the ribbed cuff and the sleeve seam, showing rib elasticity and stitch density
下摆the hem band and side seam, showing hem width and the finishing stitch
纽扣a single button and its buttonhole, showing button material, thread cross and hole finishing
拉链the zipper teeth and puller, showing tooth pitch, metal finish and the tape stitching
刺绣 / 印花the [刺绣/印花] motif filling the frame, showing thread direction / print edge sharpness and substrate texture
织法结构the [麻花/罗纹/提花] stitch structure, showing individual yarn plies and the twist of each loop
面料纤维the fabric surface at extreme magnification, showing fibre halo and weave interlacing

每条都要补三件事:

filling the frame                      ← 特写要占满画面
shallow depth of field with the far edge softly out of focus   ← 微距的景深特征
soft directional light raking across the surface to reveal depth  ← 侧光才能显出立体纹理

4、工具调用

本技能使用 dLazy 的 gpt-image-2(图像编辑模型 + --quality high;细节图的全部价值就是纹理保真度,这是本技能唯一不能省的地方)。

调用方式

两种等价写法,选一种。统一入口会自动选后端、失败重试、建目录落盘、估算成本:

# A. 统一入口(推荐):可切任意后端,加 --dry-run 不计费空跑
node scripts/gen.mjs --task clothing-detail \
  --prompt '<见下方 Prompt 模板>' \
  --images <按下表顺序> \
  --save output/clothing-detail-<sku>.jpg

# B. 直接用 dLazy CLI(不想引入 Node 依赖时,效果等价)
dlazy gpt-image-2 --prompt '...' --images ... --save output/clothing-detail.jpg

参数约定(本技能固定用法)

参数取值理由
--images[原图];带风格参考时 [原图, 参考图]顺序即 prompt 中的 image 1 / 2
--size1024x1024(方图细节位)/ 1024x1536(竖版详情页)对应原站 1:1 / 3:4
--qualityhigh(不要降)细节图靠纹理说话
--imageFormatjpeg通用格式
--batch2取景位置有随机性
--savedocs/clothing-detail/output-<sku>-<部位>.jpg按部位归档

Command Examples

# basic call: 领口细节
dlazy gpt-image-2 \
  --prompt 'Macro detail shot for an e-commerce detail page. Zoom into the collar area of this garment and render a photorealistic close-up filling the frame. Show the ribbed crewneck collar meeting the body panel, individual yarn plies and the twist of each loop, soft directional light raking across the surface. Keep colour and stitch pattern exactly as in the source. Shallow depth of field, clean neutral background bokeh, no person, no text.' \
  --images docs/clothing-detail/garment-flatlay.jpg \
  --size 1024x1024 --quality high

# complex call: 一个 SKU 批量出 3 个部位的细节图
SRC=docs/clothing-detail/garment-flatlay.jpg
COMMON='Render a photorealistic macro close-up filling the frame. Keep the colour and stitch pattern exactly as in the source image. Shallow depth of field with the far edge softly out of focus, soft directional light raking across the surface to reveal depth, clean neutral background bokeh. No person, no text, no watermark.'
for P in collar cuff stitch; do
  case $P in
    collar) VIEW='Zoom into the collar area: the ribbed crewneck collar meeting the body panel, showing rib wale spacing and the seam join' ;;
    cuff)   VIEW='Zoom into the cuff area: the ribbed cuff and the sleeve seam, showing rib elasticity and stitch density' ;;
    stitch) VIEW='Zoom into the cable-knit panel: the stitch structure, showing individual yarn plies and the twist of each loop' ;;
  esac
  dlazy gpt-image-2 \
    --prompt "Macro detail shot for an e-commerce detail page. $VIEW. $COMMON" \
    --images "$SRC" --size 1024x1024 --quality high --imageFormat jpeg \
    --save "docs/clothing-detail/output-sku001-$P.jpg"
done

# 先估价不真跑
dlazy gpt-image-2 --dry-run --prompt '...' --images a.jpg --size 1024x1024 --quality high

延伸阅读

要查什么去哪
认证、多后端配置、输出结构、错误码references/provider-cli.md
gpt-image-2 的全部可用参数references/model-flags.md
统一入口的全部选项node scripts/gen.mjs --help

5、Prompt 模板

Macro detail shot for an e-commerce detail page.

Zoom into [部位] of this [品类 + 颜色 + 面料] and render a photorealistic close-up
that fills the frame. Show [第三节表格里的取景描述].

Keep the colour and stitch pattern exactly as in the source.

Shallow depth of field with the far edge softly out of focus,
soft directional light raking across the surface to reveal depth,
clean neutral background bokeh.

No person, no text, no watermark.

按问题追加的修正句

问题追加到 prompt 末尾
纹理像塑料Resolve individual [yarn plies / weave threads / fibre ends]; the surface must read as real textile, not plastic or CG.
放大得不够Extreme magnification: the [部位] must occupy at least 70% of the frame.
编出了不存在的工艺Do not invent any construction detail that is not visible in the source image.
整张都很实、没有微距感Only the [部位] is in focus; everything beyond [X] must fall into smooth bokeh.
颜色变了Sample the colour directly from the source image; no grading, no saturation boost.

6、执行流程

  1. 挑原图:分辨率越高越好;确认目标部位清晰可见、无遮挡。
  2. 列部位清单:一个 SKU 通常出 2~3 张(领口 + 面料 + 一个特色工艺)。
  3. 每条 prompt 只放大一个部位,从第三节取景描述抄。
  4. 补齐三件事:占满画面 / 浅景深 / 侧光。
  5. --quality high(不要降档)→ --batch 2 挑图,落盘到 docs/clothing-detail/。
  6. 质检:纹理是否真实(不是塑料感)、有没有编出不存在的工艺、颜色是否一致。

7、常见问题

现象原因处理
纹理塑料感质量档位低或原图糊--quality high + 追加解析纹理句;换高分辨率原图
放大不够,还是半身未写占比追加 70% 画面占比句
编出了原图没有的拉链/刺绣模型补全追加禁止编造句
没有微距景深未写景深追加只有目标部位对焦的句子
颜色比原图艳自动调色追加取色约束句
三个部位挤在一张图一条 prompt 写了多个部位拆成多条,一条一个部位

Tips

Visit https://dlazy.com for more information.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.