Img2img Studio

图生图工作室：先识图分析输入图片，再按风格模式生成新图。触发：用户给出/上传图片并要求「生成类似风格的新图」「换个风格重绘」「图生图」「以这张图为基础出图」「做成电商图/套图/主图/详情页」「杂志编辑风/抽象艺术/手绘插画/废片焕新」。不触发：只是「描述/分析图片」而不要求生成新图（那属于 dsh-vision-skill）。模式：photo-art（editorial 编辑杂志风 / revival 白纸手绘风 / 融合风）与 ecommerce（商品主图、卖点图、场景图、详情页、PDP、Amazon/Shopify 图包）。生图通道一级优先 ChatGPT 网页（image 2.5）：免费网页账号先用，额度用尽后自有账号（Codex/Plus）优先，再降级 API（DashScope qwen-image-3.0-pro、OpenAI gpt-image、Z.AI、Seedream、MiniMax）；支持 t2i/i2i、批量并发、质量预设。

8 starsImages, Video & Audio

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add Dogwind221/img2img-studio --skill img2img-studio

Manual install

git clone --depth 1 https://github.com/Dogwind221/img2img-studio.git /tmp/img2img-studio
mkdir -p ~/.claude/skills
cp -r /tmp/img2img-studio/. ~/.claude/skills/img2img-studio

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# 图生图工作室（img2img-studio）

> **路径约定（DSH 0.1.3+）**：本文档相对路径以**本技能资源目录**为基准解析（加载时 harness 给出 `Base directory for this skill`）；跨技能引用写 `..\dsh-vision-skill\...`（三个技能同根安装时成立）。

两层架构：

```
输入图片（附件/路径/URL）
   ↓ Layer 1: 识图（复用 dsh-vision-skill）
  结构化描述（主体/构图/色彩/材质/情绪/文字）
   ↓ 模式路由（按用户要的风格）
  photo-art（照片艺术） | ecommerce（电商）   ← 模式注册表，可扩展
   ↓ Layer 2: 生成
scripts/generate_image.mjs（多供应商自动降级）
```

## 模式注册表

| 模式 id | 用途 | 参考文档 | 路由关键词（命中即选） |
|---|---|---|---|
| `photo-art` | 照片 → 艺术化新图 | `references/photo-art-mode.md` | 杂志、编辑、editorial、抽象、艺术、手绘、插画、白纸、留白、焕新、revival、诗意、文艺、zine、拾景、拼贴、纸刊、极简海报 |
| `ecommerce` | 商品图 → 电商视觉 | `references/ecommerce-mode.md` | 主图、套图、卖点图、场景图、详情页、PDP、listing、Amazon、Shopify、TikTok、A+、广告图、社媒图、直播图、包装图、商品图、电商 |
| `chatgpt-web` | 用 ChatGPT 网页端出图（无 API key） | `references/chatgpt-web-mode.md` | chatgpt 网页、网页出图、用 ChatGPT 账号、gpt-image 网页版 |
| `image-editor` | 输入框内图片标记/抠图/涂抹擦除/改尺寸后交给供应商 | `references/image-editor.md` | 标记、标注、涂抹、擦除、抠图、移除背景、改尺寸、编辑这张图、局部重绘、image editor |

路由规则：先看用户明确说的风格词；没提时看图片内容（L1 识图判断是商品照还是生活照）；仍不确定就简短问一句（给两个选项）。用户提到的风格不在注册表里 → 按最接近的模式处理并在回复中说明假设。

**新增模式**：在 `references/` 加一个 `<mode>.md`（写清风格规范、Prompt 模板、硬约束、检查清单），再在本表加一行即可，无需改脚本。

## Step 0：路由门（通道 + 风格 + 细化需求）⛔ 必做

**任何生图任务开始前**，先按 `references/routing-gate.md` 执行路由门：用 AskUserQuestion 问清三件事，再进入对应工作流。**禁止不问直接生成。**

1. **Q1 生成通道**：展示可用通道清单，**一级主通道优先**（`node scripts/generate_image.mjs --list-providers` 看实时状态）：

Read the full skill on GitHub

Details

Skill name: img2img-studio
Repository: Dogwind221/img2img-studio
License: MIT
Repo updated: 28 October 2026

FAQ

What does the Img2img Studio skill do?

How do I install the Img2img Studio skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add Dogwind221/img2img-studio --skill img2img-studio. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Img2img Studio popular?

It lives in Dogwind221/img2img-studio, which has 8 GitHub stars and 0 forks.

Related skills

Openclaw Genpark Envoy

Alpha-Park/openclaw-genpark-envoy

Autonomous relationship management. Not just CRM, but agentic gift giving, conflict resolution, and calendar-synced hospitality execution via GenPark rails.

Images, Video & Audio

Hyperframes Motion

2boysfromtown/hyperframes-motion

Make After Effects-style motion graphics in plain HTML + GSAP — 3D camera moves, kinetic typography, cinematic transitions and four ready-to-edit 1080x1920 templates. Use when the user wants an animated product reveal, social-media ad, brand story or kinetic-text video built as a web page, or asks for "motion graphics", "animated title", "cinematic text" or a 9:16 video template.

Images, Video & Audio

Personalizer

KaramveerSidhu/personalizer

Write and edit prose in the user's voice. Use when asked to personalize or humanize writing, or learn, update, or diagnose a voice profile. Excludes code, configuration, raw data, and exact formulaic text.

Images, Video & Audio

Diagram to Image

zju-zhanglu/diagram-to-image

Convert Markdown documents containing ASCII or Unicode text diagrams into faithful draw.io PNG images. Use when the user provides a Markdown file or Markdown text and asks to extract architecture diagrams, flowcharts, or sequence diagrams from fenced or indented text blocks, generate draw.io XML through the drawio MCP, review and lint XML, gate export through browser-based human approval, and export PNGs.

Images, Video & Audio

SAAS Illustration

yanliudesign/saas-illustration

SaaS 产品英雄插画风格 —— 单一品牌强调色 + 真产品 UI 卡片 + 线稿场景。用于官网首屏、Landing Page、Feature Section、Blog Header、演讲首页，以及邻近场景（技术 talk 封面、portfolio featured project、workshop landing、research paper 顶图、产品发布邮件顶图）。把"一个真实的产品 UI 卡片"叠在"线稿场景（笔电/桌面/协作场景）"之上，配以品牌强调色的漂浮装饰（聊天气泡、小图标、logo、植物）。视觉特征：纯白底、黑色矢量线稿、单一品牌强调色（默认紫 #7C5CFF，可换任意品牌 HEX）+ 温暖肤色、干净不手绘、UI 卡越出后景边界。默认输出可直接喂给 Nano Banana / GPT-image / Midjourney / Flux 的英文提示词。关键词：SaaS illustration, SaaS hero illustration, product hero, landing page illustration, 官网插画, 产品首屏, feature illustration, hero image, purple illustration, 紫色插画, 单强调色插画, single-accent illustration, duotone illust

Images, Video & Audio

Claude 3d Harness

MAX-786/claude-3d-harness

Entry point for every 3D or Blender task - modeling, materials, lighting, cameras, rendering, animation, product shots, environments, cinematic scenes, exports. Classifies the job (fast, standard or cinematic), picks a workflow, asks the harness registry which library skills to load, then drives Blender through the harness's single Blender MCP server with render, inspect and refine checkpoints. Use it whenever the user asks to create, change, light, render, animate or export anything in 3D, even if they never say "Blender".

Images, Video & Audio

Browse the most popular Claude skills, all images, video & audio skills, or write your own with the skills generator.