---
name: picgens-gpt-image-2-prompts
description: Choose GPT Image 2 / GPT Image 2.5 visual styles and industrial prompt templates from the PicGens template library. Use when writing, rewriting, classifying, or debugging an image-generation prompt for GPT Image 2 (gpt-image-2, GPT-Image2) — 22 production templates with categories, style tags, scene tags, guidance, pitfalls, copy-ready text and JSON blocks, and linked example cases.
license: MIT
---

# PicGens GPT Image 2 Prompt Library

Turn a user's image-generation intent into a production-ready GPT Image 2 prompt using the PicGens template library: 22 industrial templates across 13 categories, each with a copy-ready text block, an agent-friendly JSON block, guidance, pitfalls, and public example cases.

## Reference

- Read `references/template-library.md` before choosing a template, style, or scene. Prefer it over memory whenever template names, categories, style tags, or template text matter.
- Every template also has a public page with example images: https://www.picgens.com/templates/<slug>
- The browsable catalog lives at https://www.picgens.com/templates
- Model notes and defaults: https://www.picgens.com/models/gpt-image-2
- Prompt-writing guide: https://www.picgens.com/guides/how-to-prompt-gpt-image-2
- Machine-readable site index: https://www.picgens.com/llms.txt

## Workflow

1. Detect the user's language and answer in that language.
2. Identify the deliverable: UI screen, infographic, poster, product shot, brand mark, architecture render, photograph, illustration, character, scene, historical reconstruction, document, or special task.
3. Match in this order — category, then visual style tag, then scene tag, then nearest example case.
4. If one template is clearly strongest, use it. If two or three are plausible, present them with one-line reasons and ask the user to choose.
5. Fill the template's placeholders (`[...]` slots) with the user's specifics. Never hand back an unfilled slot.
6. Build the final prompt from these blocks:
   - subject and task
   - composition and layout
   - visual style, materials, and lighting
   - text and label requirements (exact strings, in the exact language)
   - aspect ratio and output format
   - constraints and negative details
7. Close with the template name, its https://www.picgens.com/templates/<slug> page, and any example case links you used.

## Output defaults

- Lead with the copyable prompt. Commentary goes after it, not before.
- Keep constraints concrete: exact on-image text, aspect ratio, readable labels, layout hierarchy, artifacts to avoid.
- Use the JSON block when the caller is another agent or an API; use the text block when a human will paste it.
- Chinese request → Chinese prompt; English request → English prompt, unless the user asks otherwise.
- On-image Chinese text must be quoted verbatim in the prompt, with an explicit "no garbled or placeholder text" constraint.
- For a set of variations, hold the template fixed and vary subject, composition, palette, and scene.

## Related

- Template catalog and example images: https://www.picgens.com/templates
- Prompt-template collection: https://www.picgens.com/prompts/prompt-templates
- About this skill, install instructions, and updates: https://www.picgens.com/skill

## License and attribution

Template copy and prompt text derive from **awesome-gpt-image-2** by **freestylefly**, used under the MIT License: https://github.com/freestylefly/awesome-gpt-image-2

Packaging, editorial structure, and links are by PicGens (https://www.picgens.com). Full notice in `LICENSE.md`. Example images stay with their original creators; a link is not a licence to reuse an image.