PluginBench
Skill
Review
Audit score 70

story-cover

worldwonderer/oh-story-claudecode

AI-powered web novel cover generator with title and author name rendering

What is story-cover?

Automatically analyzes story genre and style from title and author name, then generates professional web novel covers using GPT-Image-2 with embedded text. Triggered by commands like /story-cover, /封面, or natural language requests for cover design.

  • Analyzes book title and genre to determine visual style and color palette
  • Generates complete cover designs with title and author name rendered directly on the image
  • Supports multiple platforms with automatic aspect ratio adjustment (3:4 for Tomato, 2:3 default)
  • Provides 2-3 composition variants (character closeup, full body, or scene-only) for selection
  • Automatically crops and scales output to platform specifications without distortion
  • Saves prompt copies alongside generated covers for iterative refinement

How to install story-cover

npx skills add https://github.com/worldwonderer/oh-story-claudecode --skill story-cover
Prerequisites
  • GPT_IMAGE_API_KEY environment variable (OpenAI or compatible API key)
  • BOOK_DIR environment variable pointing to output directory
  • curl, jq, and base64 command-line tools
  • Optional: GPT_IMAGE_BASE_URL for proxy endpoints, REF_IMAGE for reference-based generation
Claude Code
Cursor
Windsurf
Cline

How to use story-cover

  1. 1.Collect required information: book title, author name (pen name), target platform, and set BOOK_DIR environment variable
  2. 2.Determine genre by scanning title keywords against the cover-styles reference table
  3. 3.Construct the prompt in English combining text layer (title + author styling), style layer (platform-specific keywords), and scene layer (genre-specific visuals)
  4. 4.Call the API using either text-to-image (default) or image-to-image mode if REF_IMAGE is set
  5. 5.Verify output, review generated prompt copy, and iterate if needed by adjusting style parameters

Use cases

Good for
  • Create professional covers for web novel submissions to platforms like Tomato Novel
  • Generate multiple cover design variations to test different visual approaches
  • Quickly produce covers matching specific genre aesthetics (xianxia, urban, mystery, sci-fi, etc.)
  • Adapt existing reference images into new cover designs using image-to-image generation
Who it's for
  • Web novel authors and self-publishing writers
  • Content creators on Chinese web novel platforms
  • Publishing teams needing rapid cover iteration
  • Anyone generating covers for multiple story titles

story-cover FAQ

What if the generated cover cuts off the title or author name?

Set UPLOAD_SIZE environment variable to your platform's specifications (e.g., 600x800 for Tomato). Step 3.5 will automatically center-crop and scale the output to fit without distortion, independent of the API's actual output size.

Can I use a reference image to guide the cover style?

Yes. Set the REF_IMAGE environment variable to a local path or URL, and the skill will switch to image-to-image mode using /v1/images/edits endpoint instead of text-to-image.

How does the skill choose a genre if the title matches multiple categories?

It applies a priority order: xianxia > western fantasy > classical > contemporary > urban > mystery > sci-fi > historical > supernatural > light novel. If no keywords match, it defaults to urban.

What format does the API return and how is it saved?

GPT-Image-2 always returns base64-encoded PNG data. The skill decodes it and saves as PNG, plus a .prompt.txt copy of the prompt for future reference and iteration.

Which platforms are supported and what are their cover dimensions?

Tomato Novel (600×800, 3:4 ratio) and other platforms with 2:3 ratio (1024×1536 default). The skill adjusts GPT_IMAGE_SIZE accordingly and uses UPLOAD_SIZE for final platform-specific scaling.

Full instructions (SKILL.md)

Source of truth, from worldwonderer/oh-story-claudecode.


name: story-cover version: 1.0.0 description: "小说封面生成。根据书名、作者名自动分析题材风格,调用 GPT-Image-2 直接生成含标题和署名的专业级网文封面。触发方式:/story-cover、/封面、「帮我做个封面」「生成封面图」「做个小说封面」「封面设计」。" metadata: {"openclaw":{"requires":{"env":["GPT_IMAGE_API_KEY"],"bins":["curl","jq","base64"]},"primaryEnv":"GPT_IMAGE_API_KEY","source":"https://github.com/worldwonderer/oh-story-claudecode"}}

story-cover:小说封面生成

你是小说封面设计师。根据书名和题材,调用 GPT-Image-2 一次性生成包含书名和作者名的完整封面。

核心原则:封面是读者的第一印象,一眼传达题材和氛围。


环境变量

变量必填默认说明
GPT_IMAGE_API_KEY✅—OpenAI 或兼容代理的 API Key
GPT_IMAGE_BASE_URLhttps://api.openai.com/v1兼容代理时改这个
GPT_IMAGE_MODELgpt-image-2仅在测试新模型时覆盖
GPT_IMAGE_SIZE1024x1536目标比例提示(番茄 3:4→768x1024,默认 2:3→1024x1536)。官方 gpt-image-2 认任意 16 倍数尺寸(比例≤3:1),但很多中转代理会忽略 size、按预设返回约 2:3(已实测)——平台尺寸不靠它,由 Step 3.5 兜底
UPLOAD_SIZE—平台固定上传像素(番茄 600x800);设置后 Step 3.5 居中裁剪+缩放出上传版(不变形、不依赖出图尺寸)
BOOK_DIR✅—输出目录,建议 ./covers/<书名>
REF_IMAGE—参考图本地路径或 URL;设置后走 images/edits 图生图

备注:gpt-image-2 始终返回 base64,请求体不要带 response_format(旧 DALL-E 参数,gpt-image 系列不支持)。


生成流程

Step 1:收集信息

必填:书名、作者名(笔名)、目标平台、输出目录 BOOK_DIR(建议 ./covers/<书名>,调用前 export) 选填:参考图 REF_IMAGE(本地路径或 URL,设置后切换到图生图)、风格偏好、尺寸

书名和笔名是封面必需信息:缺任一必须先用 AskUserQuestion 问用户补全,不得编造或留空。

按目标平台定封面尺寸:番茄上传 600×800 是 3:4(不是 2:3),出图比例不对、平台二次裁剪就会切掉书名/笔名。

平台上传尺寸比例生成 GPT_IMAGE_SIZE(尽量)
番茄小说600×8003:4768x1024
其他平台(默认竖版)按平台规格2:31024x1536

export GPT_IMAGE_SIZE 给目标比例(官方按它出图,很多代理会忽略、返回约 2:3);平台有固定上传像素再 export UPLOAD_SIZE(番茄 600x800)。平台尺寸最终由 Step 3.5 居中裁剪+缩放保证,不依赖代理认不认 size。 平台与题材风格见 references/cover-styles.md。

Step 1.5:题材判定

扫描书名(必要时简介)中的关键词,对照 references/cover-styles.md 的「题材推断规则」表选定题材。

  • 单题材命中 → 直接采用
  • 多题材命中 → 按优先级取一:仙侠 > 西幻 > 古言 > 现言 > 都市 > 悬疑 > 科幻 > 历史 > 灵异 > 轻小说
  • 零命中 → 默认 都市

Step 2:构建提示词

提示词 = 文字层 + 风格层 + 画面层,全部用英文编写。

文字层:书名 + 作者名字体设计

在提示词中直接包含中文书名和作者名,GPT-Image-2 可直接渲染。重点描述字体风格:

Title text '书名' at top center in [书名字体风格].
Author name '作者名' at bottom center in [作者名字体风格].

书名字体风格

题材描述关键词
玄幻/仙侠bold golden brush calligraphy with metallic glow and sharp strokes
都市modern bold sans-serif with metallic silver finish
古言/宫斗elegant golden traditional Kai script with ornate decoration
现言/甜宠soft rounded handwritten style in white with pink glow
悬疑/推理distorted bold cracked letters in blood red
科幻/末世neon glowing futuristic font in electric blue
西幻metallic embossed fantasy lettering with glow effect
历史/军事heavy stone-carved seal script in deep red
灵异/恐怖eerie dripping handwritten font in sickly green
轻小说colorful cartoon outlined bubbly font

作者名字体风格(重点:作者名必须精心设计,不能只是"小字")

作者名虽小,但是封面专业感的关键。必须指定:字体 + 颜色 + 装饰元素,让作者名与书名风格呼应但不抢焦点。

题材作者名风格提示词
玄幻/仙侠small refined white serif text with faint golden glow, flanked by delicate cloud-scroll ornaments on both sides, resting on a thin horizontal gold line
都市small clean white modern text with subtle drop shadow, positioned above a thin silver horizontal divider line
古言/宫斗small elegant dark red traditional text inside a thin golden rectangular border frame with corner decorations
现言/甜宠small soft pink-white handwritten text with a tiny heart motif on the left side, light sparkle effect
悬疑/推理small pale grey text with slight blur effect, almost hidden in the shadows, a thin cracked line underneath
科幻/末世small crisp white monospace text with subtle cyan scanline overlay, flanked by small geometric brackets
西幻small bronze medieval script text with aged parchment texture, enclosed in a small decorative shield or banner shape
历史/军事small dignified white Song typeface text above a double horizontal line in dark red
灵异/恐怖small faded grey-green text slightly tilted, with a thin dripping ink line above
轻小说small playful rounded white text with pastel color outline, tiny star decorations on both sides

作者名通用规则:

  • 大小:small(不能太大抢书名焦点,也不能太小看不清)
  • 位置:at bottom center,与画面底部保持适当间距
  • 必须有装饰元素:线条/边框/小图标/光效中至少一种
  • 颜色与背景形成对比但不刺眼

风格层:平台风格

平台风格的描述关键词统一来自 references/cover-styles.md 的「平台风格」节,按目标平台直接取对应关键词串使用,不在本文件维护副本以免与参考文件漂移。

画面层:题材 + 构图

从 references/cover-styles.md 读取题材对应的风格标签、色彩、人物、背景描述。

构图变体(首次输出 2-3 个方案):

方案构图适合题材
A人物特写 + 场景全题材通用
B全身像 + 动态姿势玄幻、都市、西幻
C纯场景/氛围图悬疑、科幻、历史

完整提示词模板

Chinese web novel cover design, [平台风格].
Title text '{书名}' at top center in [书名字体风格].
Author name '{作者名}' at bottom center in [作者名字体风格 — 从上表选择].
[题材风格标签]. [人物描述]. [背景描述].
[色彩指令]. [光效指令].
Professional book cover, high detail digital painting, portrait [平台比例:番茄=3:4,默认=2:3] ratio, keep title and author name inside the central safe area away from edges (inner ~85%), no watermark

提示词技巧(实测验证)

  • 人物描述越具体越好:服饰、姿态、发型、表情、道具每个维度都指定
  • 背景分层:前景(人物)→ 中景(场景)→ 远景(氛围)
  • 光效是指定光源方向 + 颜色(如 dramatic golden light from above)
  • 用 digital painting style 而非 photo,避免真人照片感

Step 3:调用 API 并保存

gpt-image-2 始终返回 base64,请求体不要带 response_format。$PROMPT 为 Step 2 拼出的完整提示词。

两种调用方式二选一:未设置 REF_IMAGE → 走「文生图」;设置了 → 走「图生图」。

文生图(默认)

set -euo pipefail
: "${GPT_IMAGE_API_KEY:?请设置 export GPT_IMAGE_API_KEY=你的key}"
: "${PROMPT:?请先 export PROMPT=Step 2 拼好的完整提示词}"
BASE_URL="${GPT_IMAGE_BASE_URL:-https://api.openai.com/v1}"
MODEL="${GPT_IMAGE_MODEL:-gpt-image-2}"
SIZE="${GPT_IMAGE_SIZE:-1024x1536}"
BOOK_DIR="${BOOK_DIR:?请先 export BOOK_DIR=./covers/<书名>}"

mkdir -p "$BOOK_DIR/封面"

# 自增版本号,避免覆盖之前生成的封面
i=1
while [ -f "$BOOK_DIR/封面/封面_v${i}.png" ]; do i=$((i+1)); done
OUT="$BOOK_DIR/封面/封面_v${i}.png"
RESP=$(mktemp)
trap 'rm -f "$RESP"' EXIT

# 用 jq 拼 JSON 体,避免 PROMPT 里的引号/换行/中文把 shell 字符串撑破
BODY=$(jq -n \
  --arg m "$MODEL" \
  --arg p "$PROMPT" \
  --arg s "$SIZE" \
  '{model:$m, prompt:$p, size:$s}')

curl -fsS --max-time 180 --retry 2 --retry-delay 5 \
  "$BASE_URL/images/generations" \
  -H "Authorization: Bearer $GPT_IMAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d "$BODY" > "$RESP"

# API 出错时早退,避免把 error JSON 当成 base64 写成损坏 PNG
if jq -e '.error' "$RESP" >/dev/null 2>&1; then
  echo "API error:" >&2
  jq '.error' "$RESP" >&2
  exit 1
fi

# `// empty` 让缺失字段输出空串而非 "null",配合下面的 -s 检查避免写出 3 字节假 PNG
jq -er '.data[0].b64_json // empty' "$RESP" | base64 --decode > "$OUT"
[ -s "$OUT" ] || { echo "empty or malformed output: $OUT" >&2; head -c 300 "$RESP" >&2; exit 1; }

# 落地提示词副本,方便迭代时基于上一次微调
printf '%s\n' "$PROMPT" > "${OUT%.png}.prompt.txt"

file "$OUT"
ls -lt "$BOOK_DIR/封面/"

图生图(提供参考图时)

/v1/images/edits 走 multipart/form-data,不能 用 Content-Type: application/json。文本字段用 --form-string(避免 @ 被误判为文件引用),图片字段用 -F image=@path。

set -euo pipefail
: "${GPT_IMAGE_API_KEY:?请设置 export GPT_IMAGE_API_KEY=你的key}"
: "${PROMPT:?请先 export PROMPT=Step 2 拼好的完整提示词}"
BASE_URL="${GPT_IMAGE_BASE_URL:-https://api.openai.com/v1}"
MODEL="${GPT_IMAGE_MODEL:-gpt-image-2}"
SIZE="${GPT_IMAGE_SIZE:-1024x1536}"
BOOK_DIR="${BOOK_DIR:?请先 export BOOK_DIR=./covers/<书名>}"
REF_IMAGE="${REF_IMAGE:?请先 export REF_IMAGE=本地路径或 URL}"

mkdir -p "$BOOK_DIR/封面"

# 自增版本号
i=1
while [ -f "$BOOK_DIR/封面/封面_v${i}.png" ]; do i=$((i+1)); done
OUT="$BOOK_DIR/封面/封面_v${i}.png"
RESP=$(mktemp)
REF_TMP=""
trap '[ -n "$REF_TMP" ] && rm -f "$REF_TMP"; rm -f "$RESP"' EXIT

# URL 先下载到临时文件,本地路径直接用。用裸 mktemp 以保证 macOS/Linux 行为一致。
case "$REF_IMAGE" in
  http://*|https://*)
    REF_TMP=$(mktemp)
    curl -fsSL --max-time 60 -o "$REF_TMP" "$REF_IMAGE"
    REF_LOCAL="$REF_TMP"
    ;;
  *)
    [ -f "$REF_IMAGE" ] || { echo "参考图不存在: $REF_IMAGE" >&2; exit 1; }
    REF_LOCAL="$REF_IMAGE"
    ;;
esac

curl -fsS --max-time 240 --retry 2 --retry-delay 5 \
  "$BASE_URL/images/edits" \
  -H "Authorization: Bearer $GPT_IMAGE_API_KEY" \
  --form-string "model=$MODEL" \
  --form-string "size=$SIZE" \
  --form-string "prompt=$PROMPT" \
  -F "image=@$REF_LOCAL" > "$RESP"

if jq -e '.error' "$RESP" >/dev/null 2>&1; then
  echo "API error:" >&2
  jq '.error' "$RESP" >&2
  exit 1
fi

# `// empty` 让缺失字段输出空串而非 "null",配合 -s 检查避免写出 3 字节假 PNG
jq -er '.data[0].b64_json // empty' "$RESP" | base64 --decode > "$OUT"
[ -s "$OUT" ] || { echo "empty or malformed output: $OUT" >&2; head -c 300 "$RESP" >&2; exit 1; }

printf '%s\n' "$PROMPT"    > "${OUT%.png}.prompt.txt"
printf '%s\n' "$REF_IMAGE" > "${OUT%.png}.ref.txt"

file "$OUT"
ls -lt "$BOOK_DIR/封面/"

Step 3.5:导出平台上传尺寸(平台有固定像素时)

设了 UPLOAD_SIZE(番茄 600×800)就把原图居中裁剪+缩放成上传尺寸——不论出图是 2:3 还是 3:4 都裁成平台精确像素,不变形,避免平台再裁切掉书名/笔名。原图保留、另存 _上传 版:

SRC="${OUT:-$(ls -t "${BOOK_DIR:-.}"/封面/封面_v*.png 2>/dev/null | grep -v _上传 | head -1)}"  # 复用 Step 3 的 $OUT;新 shell 里从 BOOK_DIR 找最新原图
TARGET="${UPLOAD_SIZE:-}"   # 番茄=600x800;未设则跳过
if [ -n "$TARGET" ] && [ -f "$SRC" ]; then
  UP="${SRC%.png}_上传.png"; W="${TARGET%x*}"; H="${TARGET#*x}"
  if command -v magick >/dev/null 2>&1; then M=magick
  elif command -v convert >/dev/null 2>&1; then M=convert; else M=""; fi
  if [ -n "$M" ]; then
    "$M" "$SRC" -resize "${W}x${H}^" -gravity center -extent "${W}x${H}" "$UP"  # 缩放填满后居中裁
  elif command -v sips >/dev/null 2>&1; then
    cp "$SRC" "$UP"
    sw=$(sips -g pixelWidth "$UP" | awk '/pixelWidth/{print $NF}')
    sh=$(sips -g pixelHeight "$UP" | awk '/pixelHeight/{print $NF}')
    if [ $((sw*H)) -ge $((sh*W)) ]; then sips --resampleHeight "$H" "$UP" >/dev/null
    else sips --resampleWidth "$W" "$UP" >/dev/null; fi
    sips -c "$H" "$W" "$UP" >/dev/null   # sips -c 是 高 宽,居中裁
  else
    echo "无 magick/convert/sips,跳过;手动把 $SRC 居中裁剪+缩放到 $TARGET 再上传" >&2
  fi
  [ -f "$UP" ] && file "$UP"
fi

书名/笔名已在提示词里留中心安全区,居中裁剪不会切到。

Step 4:质量检查 + 迭代

检查项标准
文字渲染书名清晰可辨,字体风格匹配题材
题材匹配视觉风格与书名题材一致
构图合理主体突出,文字不遮挡核心画面
平台适配符合目标平台的封面风格调性
平台尺寸比例与平台一致;缩放到上传尺寸后书名、笔名完整可见、未被裁切

不满意时调整方向:更换构图、调整色调、换字体风格、换平台风格。


参考资料

文件何时加载
references/cover-styles.md题材→视觉风格映射、平台风格详情、提示词模板

语言

  • 跟随用户的语言回复,用户用什么语言就用什么语言回复
  • 中文回复遵循《中文文案排版指北》