# GPT Image Hub

> GPT Image Hub is a prompt library and image generation workspace for GPT Image 2 and compatible image models.

The library has 1516 public prompt templates with full prompt text, model defaults, aspect ratio hints, and source attribution. The site is available in 18 languages, and prompts can be used in the web generator or through the authenticated public API.

This file is an auxiliary discovery guide. Inclusion here does not guarantee search ranking, indexing, or citation by an AI system.

## Core pages

- [Prompt library](https://www.gptimagehub.com/): browse and search every public prompt template by category.
- [Image generator](https://www.gptimagehub.com/generate): generate images from a prompt template in the browser.
- [AI image model guides](https://www.gptimagehub.com/models): what GPT Image 2 and the Nano Banana models are best at, with supported settings.
- [Model comparison](https://www.gptimagehub.com/models/compare): capability matrix and decision rules across the supported image models.
- [Prompt engineering guides](https://www.gptimagehub.com/guides): practical guides for writing reusable image prompts.
- [Pricing](https://www.gptimagehub.com/pricing): credit packs, prices, and what each plan includes.
- [About](https://www.gptimagehub.com/about): what the site provides, who it is for, and how it is maintained.
- [Editorial policy](https://www.gptimagehub.com/editorial-policy): prompt quality standards, safety and rights, updates and corrections.
- [Dataset methodology](https://www.gptimagehub.com/prompt-dataset-methodology): dataset scope, formats, provenance, and freshness.

## Prompt categories

- [3D Render](https://www.gptimagehub.com/library/category/3d-render): 17 prompts.
- [Anime / Manga](https://www.gptimagehub.com/library/category/anime-manga): 31 prompts.
- [Character Design](https://www.gptimagehub.com/library/category/character-design): 36 prompts.
- [Cinematic / Film Still](https://www.gptimagehub.com/library/category/cinematic): 37 prompts.
- [Concept Art](https://www.gptimagehub.com/library/category/concept-art): 15 prompts.
- [Game Asset](https://www.gptimagehub.com/library/category/game-asset): 25 prompts.
- [Infographic / Education](https://www.gptimagehub.com/library/category/infographic): 97 prompts.
- [Logo & Branding](https://www.gptimagehub.com/library/category/logo-branding): 21 prompts.
- [Other](https://www.gptimagehub.com/library/category/other): 61 prompts.
- [Portrait & Photography](https://www.gptimagehub.com/library/category/portrait-photography): 815 prompts.
- [Poster & Illustration](https://www.gptimagehub.com/library/category/poster-illustration): 219 prompts.
- [Product & E-commerce](https://www.gptimagehub.com/library/category/product-ecommerce): 49 prompts.
- [Social Media](https://www.gptimagehub.com/library/category/social-media): 35 prompts.
- [UI / UX Design](https://www.gptimagehub.com/library/category/ui-ux-design): 58 prompts.

## Prompt engineering guides

- [Best GPT Image Prompts for Product Photography](https://www.gptimagehub.com/guides/best-gpt-image-prompts-for-product-photography): The best GPT Image product photography prompts specify the exact product, surface, background, lighting, camera angle, composition, brand-safe text, and intended commerce use case. Strong prompts also separate product facts from style direction so agents can adapt them without changing the product identity.
- [How to Write Consistent Character Prompts](https://www.gptimagehub.com/guides/consistent-character-prompts): Consistent character prompts describe a stable character bible: age range, silhouette, face shape, hair, clothing, palette, personality cues, recurring props, and the exact output format. The prompt should also state which details may change between poses and which details must remain fixed.
- [GPT Image Prompt Examples by Style](https://www.gptimagehub.com/guides/gpt-image-prompt-examples-by-style): A good style prompt names the target visual family, then adds medium, composition, palette, lighting, texture, camera, and output use case. Style words are most useful when paired with concrete constraints such as line weight, lens, typography, material, or layout grid.
- [Midjourney vs GPT Image Prompts](https://www.gptimagehub.com/guides/midjourney-vs-gpt-image-prompts): Midjourney prompts often behave like compact visual mood boards, while GPT Image prompts can support more explicit instruction following, structured JSON-like fields, exact layout requirements, and workflow-ready templates. GPT Image prompts should usually be clearer about task, constraints, text, and edit intent.
- [How to Write GPT Image Prompts for UI Mockups](https://www.gptimagehub.com/guides/gpt-image-ui-mockup-prompts): Good GPT Image prompts for UI mockups name the product, the screen and its main components, the device or frame it appears in, the visual style, and the exact text that must be readable. List components in reading order and keep on-screen copy short, because the model renders a few large labels far more reliably than paragraphs of interface text.
- [How to Write Poster Prompts for GPT Image 2](https://www.gptimagehub.com/guides/gpt-image-poster-prompts): Effective poster prompts define the poster's purpose, a single focal image, the composition and negative space, the exact title and supporting text with their placement, the illustration or photo style, the palette, and the aspect ratio. Posters fail most often on crowded composition and long text, so state where the empty space goes and keep typography to a title and one or two short lines.
- [How to Write Logo and Brand Identity Prompts](https://www.gptimagehub.com/guides/gpt-image-logo-branding-prompts): Logo and brand prompts work best when they state the brand name exactly, what the business does, the mark type (wordmark, lettermark, symbol, or emblem), two or three personality traits, color limits, and the presentation format. For a full identity board, list the deliverables, such as logo, typography, palette, and social templates, so the model lays them out as a brand guideline page.

## Image model guides

- [GPT Image 2 model guide](https://www.gptimagehub.com/models/gpt-image-2): GPT Image 2 is the best default choice on GPT Image Hub when you need high-quality generation or editing, many aspect-ratio choices, and up to 16 reference images. Use it for polished commercial visuals, product mockups, editorial assets, and prompts where instruction following matters more than the lowest possible iteration latency.
- [GPT Image 2 Saver model guide](https://www.gptimagehub.com/models/gpt-image-2-saver): GPT Image 2 Saver runs OpenAI's gpt-image-2 through a lower-cost provider route that bills per image, so GPT Image Hub charges a flat 5 credits (about $0.005) whether you pick low, medium, or high quality at 1K, 2K, or 4K. In the Hub's tests it took 23 to 43 seconds per image, rendered on-image text accurately, and handled multi-image reference edits. Choose it for high-volume drafts, social content, and trying prompts cheaply; choose standard GPT Image 2 when you need up to 16 reference images.
- [GPT Image 2.5 Flare model guide](https://www.gptimagehub.com/models/gpt-image-2-5-flare): GPT Image 2.5 Flare is OpenAI's default GPT Image 2.5 model for fast, high-quality everyday image generation. OpenAI positions it above GPT Image 2 in quality at roughly half the latency, billed at the same token rates; in GPT Image Hub's own tests a 1536×1024 image at high quality used 1,372 output tokens and 27 seconds, against 5,488 tokens and 78 seconds for GPT Image 2 at the same size. Choose it for most new work; choose Sunburst when editing precision matters most.
- [GPT Image 2.5 Sunburst model guide](https://www.gptimagehub.com/models/gpt-image-2-5-sunburst): GPT Image 2.5 Sunburst is the GPT Image 2.5 model OpenAI recommends when precision matters most. It supports the same xhigh and max quality settings, custom resolutions, and transparent backgrounds as Flare, and GPT Image Hub runs it at xhigh by default; in Hub tests it took 47–108 seconds per image and placed every label and line of text exactly as prompted. Choose it for packaging, typography-heavy layouts and intricate illustration; choose Flare when speed matters more.
- [Nano Banana model guide](https://www.gptimagehub.com/models/nanobanana): Nano Banana is the fast Gemini image option on GPT Image Hub. Use it when you want quick creative iteration, light image edits, and reference-image guided outputs at 1K. It is a good exploration model before moving a winning prompt into GPT Image 2, Nano Banana 2, or Nano Banana Pro for larger exports.
- [Nano Banana 2 model guide](https://www.gptimagehub.com/models/nanobanana2): Nano Banana 2 is the balanced Gemini choice on GPT Image Hub. It supports 1K, 2K, and 4K tiers in the app, keeps the fast Flash-style workflow, and allows up to 14 reference images. Choose it when Nano Banana is too limited but Nano Banana Pro is more than the task needs.
- [Nano Banana 2.1 model guide](https://www.gptimagehub.com/models/nanobanana2-1): Nano Banana 2.1 is Google's update to Nano Banana 2 and the model Google now recommends for new projects. Its image output costs half as much as Nano Banana 2, so on GPT Image Hub it is 27 credits at 1K, 41 at 2K, and 91 at 4K. In the Hub's tests it rendered a menu with every price correct and turned one concept from a reference sheet into a finished sticker. It thinks before drawing, so 2K and 4K images can take several minutes.
- [Nano Banana Pro model guide](https://www.gptimagehub.com/models/nanobananapro): Nano Banana Pro is the premium Gemini image option on GPT Image Hub. Use it for detailed, high-resolution images, complex visual reasoning, branded scenes, product concepts, and outputs where typography or real-world knowledge may matter. It supports up to 14 reference images and 1K, 2K, and 4K tiers in the app.
- [Nano Banana 2 Lite model guide](https://www.gptimagehub.com/models/nanobanana2-lite): Nano Banana 2 Lite is Google's fastest and cheapest Gemini image model. On GPT Image Hub it renders 1K images for 27 credits, and in the Hub's tests most images finished in 12 to 26 seconds. Google notes that it is not optimized for many reference images or long edit sequences, so choose Nano Banana 2.1 for text-heavy layouts, 2K or 4K output, and multi-image edits.
- [Seedream 4.0 model guide](https://www.gptimagehub.com/models/seedream-4-0): Seedream 4.0 is ByteDance's image generation and editing model. On GPT Image Hub it costs 25 credits per image at any size, 1K, 2K or 4K, which makes its 4K output one of the cheapest on the site. In the Hub's tests 1K and 2K images took about 20 seconds and a 4K landscape took 137 seconds. Choose it for photographic scenes and large exports on a budget; choose Seedream 4.5 or 5.0 for stronger text and design work.
- [Seedream 4.5 model guide](https://www.gptimagehub.com/models/seedream-4-5): Seedream 4.5 is ByteDance's image model with stronger typography and design control than 4.0. On GPT Image Hub it renders 2K and 4K images for 33 credits each; it has no 1K tier because it needs at least about 3.7 megapixels. In the Hub's tests it rendered clean poster text, replaced a headline and background in a reference edit, and produced a 4K product photo in 28 seconds.
- [Seedream 5.0 model guide](https://www.gptimagehub.com/models/seedream-5-0): Seedream 5.0 on GPT Image Hub is ByteDance's lowest-priced 5.0 model: 15 credits per image at 2K or 4K. BytePlus lists the 5.0 family as Pro, Lite, and Flash, and the Hub prices this model as 5.0 Flash. In the Hub's tests it produced a detailed 4K festival illustration in 129 seconds and a three-view character sheet with correct labels in 46 seconds.
- [Seedream 5.0 Pro model guide](https://www.gptimagehub.com/models/seedream-5-0-pro): Seedream 5.0 Pro is ByteDance's most capable Seedream model, aimed at design work with accurate typography. On GPT Image Hub it costs 37 credits at 1K and 73 at 2K, following BytePlus's price step at 2.61 megapixels; it has no 4K tier. In the Hub's tests it rendered a Kyoto festival poster with a large title and a dated subtitle exactly as prompted. It can be slow: that 2K poster took 290 seconds, while a 1K image took 46 seconds.
- [Seedream 5.0 Lite model guide](https://www.gptimagehub.com/models/seedream-5-0-lite): Seedream 5.0 Lite is the lighter model in ByteDance's Seedream 5.0 family. On GPT Image Hub it renders 2K images in eight aspect ratios, from 1:1 to 21:9, for 36 credits each, at BytePlus's full list price. In the Hub's tests it produced natural portraits and interiors in 25 to 33 seconds and relit a portrait from a reference image while keeping the person and scene.
- [Z-Image Turbo model guide](https://www.gptimagehub.com/models/z-image-turbo): Z-Image Turbo is Alibaba Tongyi's fast text-to-image model. On GPT Image Hub it renders 1K and 2K images for 15 credits each, at Alibaba's full list price, in any of the Hub's 16 aspect ratios. In the Hub's tests a 2K product photo took 11 seconds and 1K scenes took 42 to 70 seconds. It does not accept reference images, so use another model for edits.

## Agent and API access

- [Prompt search API](https://www.gptimagehub.com/api/agent/prompts): public prompt search returning full prompt text; `/api/agent/prompts/{slug}` returns one record.
- [API documentation](https://www.gptimagehub.com/api-docs): authenticated image generation API.
- [OpenAPI spec](https://www.gptimagehub.com/openapi.json): machine-readable API description.
- [MCP endpoint](https://www.gptimagehub.com/mcp): tool-based prompt discovery and generation URL creation.
- [Agent manifest](https://www.gptimagehub.com/.well-known/agent-manifest.json): index of agent-facing endpoints.
- [Agent skills](https://www.gptimagehub.com/.well-known/agent-skills/index.json): reusable agent skill definitions.
- Follow each public page’s text/markdown alternate link, or call `/api/agent/markdown?path={path}`. Page URLs always return HTML.

## Public resources

- [GitHub prompt dataset](https://github.com/jqy-yo/gptimagehub-prompts): Attributed CC BY 4.0 prompt snapshot, preview gallery, and generated examples. This licensed subset is not the complete live library.
- [Hugging Face dataset](https://huggingface.co/datasets/jqyyyy/gptimagehub-prompts): Download and explore the same licensed prompt snapshot published on GitHub. Individual image rights remain with their respective owners.
- [Official MCP Registry listing](https://registry.modelcontextprotocol.io/?q=io.github.jqy-yo%2Fgptimagehub): Registry entry for io.github.jqy-yo/gptimagehub, connecting agents to the website's public MCP tools.

## Citation guidance

- Cite canonical prompt detail URLs for prompt templates and model guide URLs for model capability questions.
- Cite the dataset manifest or JSONL export for bulk prompt corpus references.
- Do not cite private account, admin, billing, or generation history pages.

## Optional

- [Extended context](https://www.gptimagehub.com/llms-full.txt): this overview plus model capabilities and the most-used prompts in full.
- [Prompt dataset (JSON Lines)](https://www.gptimagehub.com/datasets/prompts.jsonl): every public prompt, for bulk retrieval.
- [Dataset manifest](https://www.gptimagehub.com/datasets/prompts-manifest.json): dataset version, formats, and per-category files.
- [Sitemap index](https://www.gptimagehub.com/sitemap-index.xml): every indexable URL in every language.
- [How it works](https://www.gptimagehub.com/home): how prompt templates improve GPT Image 2 results, in three steps.
- Supported locales: en, zh-CN, zh-TW, ja, ko, es, pt-BR, fr, de, it, ru, ar, hi, tr, id, vi, th, pl.