# GPT-Image2 Prompt Templates: The Two Forms Explained

> Discover the two GPT-Image2 prompt templates: human-readable plain text for easy editing and structured JSON for automation. Learn more in the awesome-gpt-image-2 repo.

- Repository: [苍何/awesome-gpt-image-2](https://github.com/freestylefly/awesome-gpt-image-2)
- Tags: deep-dive
- Published: 2026-08-26

---

**The awesome-gpt-image-2 repository defines two distinct prompt template formats—a human-readable plain-text form for manual editing and a structured JSON form for programmatic automation—both cataloged in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md).**

The freestylefly/awesome-gpt-image-2 repository maintains a master collection of GPT-Image2 prompt templates designed to generate high-quality UI screenshots, infographics, and posters. Every template category in the codebase provides dual-format definitions to support both interactive human workflows and deterministic machine processing.

## The Two GPT-Image2 Prompt Template Formats

According to the source code in `freestylefly/awesome-gpt-image-2`, every prompt template is provided in two parallel formats side-by-side within each category (UI, Infographic, Poster, and others). These are referred to as the **Plain-Text "常规模板"** (Regular Template) and the **JSON "进阶模板"** (Advanced Template).

### Plain-Text "常规模板" (Regular Template)

The plain-text format is a human-readable, free-form text template that describes the desired image in natural language with placeholders. This form uses bracketed variables such as `[产品类型]`, `[平台]`, and `[功能点A]` to indicate customizable fields.

Key characteristics include:

- Natural language structure optimized for readability
- Placeholder syntax using square brackets (e.g., `[产品类型]`, `[平台，如 iOS/Android/Web]`)
- Immediate copy-paste compatibility for direct user interaction
- Manual editing capability without parsing requirements

In the repository's [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md), the UI category's plain-text template appears at lines 16-22, immediately followed by its JSON counterpart at lines 24-46. The same pattern repeats for the Infographic section (lines 98-105 for plain-text, lines 107-129 for JSON) and the Poster section (lines 51-58 for plain-text, lines 60-68 for JSON).

### JSON "进阶模板" (Advanced Template)

The JSON format is a structured object that enumerates each component of the prompt with explicit type fields. This machine-readable structure is optimized for programmatic consumption by agents, automation scripts, and API integrations.

Key characteristics include:

- Strict schema with named fields: `type`, `platform`, `product`, `layout`, `style`, `content`, `constraints`
- Nested objects for complex properties (e.g., `style.theme`, `content.cards`)
- Deterministic field extraction for batch processing
- Direct compatibility with automation agents and API payloads

The JSON templates in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md) serve as the canonical reference for the [`agents/skills/gpt-image-2-style-library/references/style-library.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/references/style-library.md) file, which links directly to these structured definitions for programmatic use.

## Source File Locations

The dual-format templates are distributed across three key files in the repository:

### docs/templates.md

This file contains the master catalog of all prompt templates, showing both the plain-text and JSON forms for every category. It serves as the primary reference document where both formats appear side-by-side for comparison.

### agents/skills/gpt-image-2-style-library/references/style-library.md

This file provides template references used specifically by the `gpt-image-2` skill, linking directly to the JSON versions in [`templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/templates.md). It enables agent-based automation workflows to consume the structured templates programmatically.

### README.md

The main documentation file (and its localized variants) links to the template sections and explains the semantic differences between the two forms, guiding users on when to select each format.

## Practical Examples: UI Category

The following examples from [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md) demonstrate the identical UI screenshot prompt expressed in both formats.

**Plain-Text Template (Lines 16-22):**

```text
为[产品类型]生成一张[平台，如 iOS/Android/Web]界面图。
核心功能：[功能点A]、[功能点B]、[功能点C]。
视觉风格：[极简/科技/拟物]，主色[颜色]，强调色[颜色]。
布局：[顶部导航/双栏/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[9:16/16:9]。

```

**JSON Template (Lines 24-46):**

```json
{
  "type": "UI Screenshot",
  "platform": "iOS",
  "product": "Fitness App",
  "layout": "Card-based feed with bottom tab bar",
  "style": {
    "theme": "Dark Mode",
    "primary_color": "Neon Green",
    "typography": "Clean sans-serif"
  },
  "content": {
    "header": "Today's Activity",
    "cards": [
      {"title": "Running", "data": "5.2 km", "button": "Start"},
      {"title": "Calories", "data": "340 kcal"}
    ]
  },
  "constraints": "High fidelity, readable text, 9:16 aspect ratio"
}

```

## When to Use Each Format

Select the appropriate template form based on your workflow requirements:

- **Plain-Text Form**: Use when engaging in direct user interaction, quick copy-paste operations, or manual editing scenarios where human readability takes precedence over structure. This format allows rapid iteration without JSON syntax constraints.

- **JSON Form**: Use for agent-based automation, batch processing workflows, or API integrations requiring deterministic field extraction. This format enables scripts to generate variations programmatically by modifying specific object properties without parsing natural language.

## Summary

- The repository maintains **two parallel formats** for every GPT-Image2 prompt template: plain-text "常规模板" and JSON "进阶模板".
- **Plain-text templates** in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md) use bracketed placeholders like `[产品类型]` for human-readable, editable prompts at lines 16-22 (UI), 51-58 (Poster), and 98-105 (Infographic).
- **JSON templates** provide structured schemas with fields like `type`, `layout`, `style`, and `constraints` at lines 24-46 (UI), 60-68 (Poster), and 107-129 (Infographic) for machine consumption.
- The [`agents/skills/gpt-image-2-style-library/references/style-library.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/references/style-library.md) file references the JSON templates for programmatic agent use.
- Choose plain-text for manual workflows and JSON for automation scripts requiring deterministic data extraction.

## Frequently Asked Questions

### What is the difference between the plain-text and JSON GPT-Image2 prompt templates?

The plain-text template is a human-readable format using natural language with bracketed placeholders like `[产品类型]`, while the JSON template is a structured object with explicit fields such as `type`, `platform`, and `style`. According to the source code in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md), both contain identical semantic information but differ in syntax optimized for human editing versus machine parsing.

### Which template format should I use for automation scripts?

Use the **JSON "进阶模板"** format for automation scripts. This structured format appears in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md) and is referenced by [`agents/skills/gpt-image-2-style-library/references/style-library.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/references/style-library.md) specifically for agent consumption. The JSON schema allows deterministic extraction of fields like `layout`, `content`, and `constraints` without natural language parsing.

### Where are the GPT-Image2 prompt templates located in the repository?

All templates reside in [`docs/templates.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/templates.md), which catalogs both formats for every category (UI, Infographic, Poster). The JSON versions are additionally linked from [`agents/skills/gpt-image-2-style-library/references/style-library.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/references/style-library.md) for programmatic access, while [`README.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/README.md) provides high-level documentation linking to both forms.

### Can I convert between the plain-text and JSON formats?

While the repository presents both formats side-by-side (e.g., UI plain-text at lines 16-22 and JSON at lines 24-46), manual conversion requires mapping bracketed placeholders to JSON fields. The plain-text placeholders like `[平台，如 iOS/Android/Web]` correspond to JSON keys such as `"platform"`, but automated conversion utilities are not provided in the current codebase.