GPT-Image-2 Industrial Prompt Templates: Complete Guide to All 21 Categories

GPT-Image-2 provides 21 industrial-grade prompt templates covering UI design, infographics, posters, products, branding, architecture, photography, illustration, characters, and narrative scenes—each with built-in "pitfalls-avoidance" guidelines for production-ready outputs.

The freestylefly/awesome-gpt-image-2 repository ships a curated industrial prompt template system (工业级提示词模板) that transforms rough concepts into structured, reproducible image-generation prompts. These templates live in [docs/templates.md](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md) and are designed for API-driven agents, batch workflows, and consistent commercial output.

What Are Industrial Prompt Templates?

Industrial prompt templates in GPT-Image-2 are pre-engineered prompt patterns that encode domain expertise. Unlike ad-hoc prompts, each template specifies:

• Input structure – what variables to provide (product name, style keywords, constraints)
• Output specification – exact format, aspect ratio, fidelity requirements
• Pitfalls-avoidance rules – explicit negative constraints (防坑指南) that prevent common failure modes

The templates split into two formats: plain-text for direct UI use and JSON-structured for programmatic agents.

All 21 Industrial Prompt Templates

Below are the complete GPT-Image-2 industrial prompt templates as documented in docs/templates.md.

UI & Interface Templates (1–4)

| # | Template | Core Pattern |

|---|----------|--------------| | 1 | UI & Interface | 为[product_type]生成一张[platform]界面图。核心功能：[features]。视觉风格：[style]。布局：[layout]。输出：[constraints] | | 2 | JSON Advanced UI | Structured JSON with type: "UI Screenshot", platform, product, layout, style, content, constraints | | 3 | Screenshot Generation | 生成一张[platform]内容截图，[mode]模式。核心内容：[elements]。约束：文字必须准确显示指定的[language] | | 4 | Live Stream Interface | 生成一张[platform]直播界面截图。主播[streamer_info]。背景[background]。UI叠加层：[overlays]。约束：文字清晰、弹幕合理、界面不遮挡主播 |

Infographic & Visualization Templates (5–6)

| # | Template | Core Pattern |

|---|----------|--------------| | 5 | Infographic & Visualization | 生成[topic]信息图，目标读者为[audience]。结构：标题区 + 3-5模块（图标、短标题、说明）。输出：层级清晰、可读性高的[language]信息图 | | 6 | JSON Advanced Infographic | JSON with type: "Infographic", topic, audience, structure, style, constraints |

Poster & Typography Templates (7–12)

| # | Template | Core Pattern |

|---|----------|--------------| | 7 | Poster & Layout | 设计一张[category]海报，主题为[theme]。主视觉[visual]。版式[layout]。色彩[color]。氛围[mood]。输出：高分辨率社媒海报 | | 8 | Sports Campaign Poster | 设计一张[sport]商业Campaign海报。主体[subject]。视觉风格[style]。约束：主体清晰，文字可读，色调统一 | | 9 | Conceptual Typography Poster | Create ONE finished premium conceptual typography poster for the exact title: "[title]". Single poster only, no alternatives, no moodboard | | 10 | Multi-Style Signature Selection Poster | 基于姓名自动生成9:16竖版多风格签名选择海报，提供6种签名方案（极简、狂放、松弛、东方行草、锋利、实验） | | 11 | Single Signature Extraction | 从输入图的[location]签名里，提取核心笔势，生成纯签名图。约束：只保留签名主体，背景纯白，墨色深黑 | | 12 | Signature Practice Breakdown | 基于输入的签名图，生成签名练习拆解图，包含8-12步骤、笔画箭头、起止点标注 |

Product & E-Commerce Templates (13–14)

| # | Template | Core Pattern |

|---|----------|--------------| | 13 | Product & E-Commerce | 生成[product_name]电商主图，卖点为[selling_points]。场景[scene]。附加元素[elements]。输出：电商平台可直接使用的商品展示图 | | 14 | Beauty Product Recommendation Report | 基于用户自拍与品牌，生成竖版口红推荐报告信息图，包含肤色分析、气质匹配、试色矩阵、品牌调性 |

Brand & Logo Templates (15–16)

| # | Template | Core Pattern |

|---|----------|--------------| | 15 | Brand & Logo | 为[brand_name]设计品牌视觉方案。关键词[keywords]。包含Logo方向、配色、字体、应用示意 | | 16 | Complete Brand Identity Package | 生成3-5个Logo方向、配色系统、字体系统、应用触点（名片、App图标等）以及品牌规则文档 |

Visual Production Templates (17–21)

| # | Template | Core Pattern |

|---|----------|--------------| | 17 | Architecture & Spatial Design | 生成[space_type]设计效果图，功能定位为[function]。风格[style]。材质[materials]。输出：写实建筑空间渲染图 | | 18 | Photography & Realism | 拍摄主题[subject]。场景[scene]。摄影参数风格[params]。光线[lighting]。输出：高写实摄影风格图像 | | 19 | Illustration & Art | 创作[theme]插画，主角为[character]。画风[style]。配色[palette]。输出：高质量插画 | | 20 | Character & Role Design | 设计[role_identity]角色设定图。外观[appearance]、性格[personality]、姿态[pose]、世界观[worldview]、标志性元素[signature_items]。输出：角色主视图 | | 21 | Scene & Narrative | 生成[story_theme]场景图，发生在[time_location]。主事件[event]。主角[protagonist]。镜头语言[camera_language]。氛围[atmosphere]。输出：具备叙事张力的场景概念图 |

Template Variants: Plain-Text vs. JSON

Each industrial prompt template offers two implementation variants.

Plain-text template – paste directly into GPT-Image-2:

为电商购物APP生成一张iOS界面图。
核心功能：商品搜索、购物车、支付。
视觉风格：极简，主色蓝色，强调色橙色。
布局：顶部导航 + 卡片流，信息层级清晰，留白充足。
输出：高保真UI截图，文字清晰可读，比例9:16。

JSON template – for agent programmatic control:

{
  "type": "UI Screenshot",
  "platform": "iOS",
  "product": "E-commerce Shopping App",
  "layout": "Card-based feed with bottom tab bar",
  "style": {
    "theme": "Minimalist",
    "primary_color": "Blue",
    "accent_color": "Orange",
    "typography": "Clean sans-serif"
  },
  "content": {
    "header": "Search, Cart, Checkout",
    "cards": [
      {"title": "Featured Product", "image": "url", "price": "$99"},
      {"title": "New Arrival", "image": "url", "price": "$59"}
    ]
  },
  "constraints": "High fidelity, readable text, 9:16 aspect ratio"
}

Pitfalls-Avoidance Guidelines (防坑指南)

Every industrial prompt template includes a pitfalls-avoidance section that specifies what the model must not do. Common rules across templates:

• Text accuracy – Chinese characters must display exactly as specified, no placeholder gibberish
• No moodboards – output finished production assets, not reference collections
• Single output – one definitive result unless explicitly requesting variants
• Clear hierarchy – no visually cluttered or ambiguous compositions
• Proper aspect ratio – strict enforcement of 9:16, 16:9, 1:1, etc.

These constraints are encoded in both the plain-text 约束 clauses and JSON constraints fields.

Repository Structure for Template Access

File Path	Purpose
docs/templates.md	Master documentation of all 21 industrial prompt templates with full patterns and pitfalls-avoidance guides
agents/skills/gpt-image-2-style-library/SKILL.md	Agent skill definition for template selection and rendering
data/style-library.json	Structured machine-readable template data
README.md	Project overview and quick-start links

Summary

• GPT-Image-2 ships 21 industrial prompt templates spanning UI, infographics, posters, typography, products, branding, architecture, photography, illustration, characters, and narrative scenes
• Two formats per template: plain-text for manual use, JSON for programmatic agents
• Built-in pitfalls-avoidance ensures production-ready, reproducible outputs without common failure modes
• Primary source: docs/templates.md in freestylefly/awesome-gpt-image-2

Frequently Asked Questions

What makes these templates "industrial-grade"?

The GPT-Image-2 industrial prompt templates are engineered for reproducibility and scale. They include explicit input schemas, output specifications, and negative constraints (pitfalls-avoidance) that prevent the ambiguous, inconsistent results typical of ad-hoc prompting. This makes them suitable for API-driven batch generation and commercial workflows.

How do I use the JSON template variants?

Parse data/style-library.json or construct the JSON structure shown in each template's "JSON Advanced" section. Pass this to GPT-Image-2 through the API or agent framework described in agents/skills/gpt-image-2-style-library/SKILL.md. The JSON format enables dynamic variable substitution and automated template selection.

Where can I find examples for a specific template category?

Each of the 21 templates in docs/templates.md includes a complete working example with all variables filled. The examples demonstrate both the plain-text and JSON formats where applicable. Copy the example, replace the bracketed placeholders with your values, and execute.

Can I modify these templates for custom use cases?

Yes. The templates are documented as patterns with clear variable slots (marked with [brackets]). Modify the constraints section to adjust aspect ratios, fidelity requirements, or platform specifications. For systematic customization, fork the repository and edit data/style-library.json to add new template variants.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}