GPT Image 2 (model id: gpt-image-2) is OpenAI's image generation model exposed through the Images API. It supports text-to-image and image editing with strong text rendering, photoreal portraits, and structured layouts like posters and UI mockups. It is the successor to DALL·E 3 and competes with Google Nano Banana Pro on the high end.

Where do these prompts come from?

They are curated from two open-source community libraries: • EvoLinkAI/awesome-gpt-image-2-prompts (Apache-2.0) — https://github.com/EvoLinkAI/awesome-gpt-image-2-prompts • YouMind-OpenLab/awesome-gpt-image-2 (CC BY 4.0) — https://github.com/YouMind-OpenLab/awesome-gpt-image-2 Each card links back to the original author and tweet, and shows a small badge identifying which library curated it.

Can I use these prompts commercially?

Yes. The prompt text is shared by the upstream curators under permissive licenses (Apache-2.0 and CC BY 4.0). Images you generate by running these prompts on your own OpenAI account belong to you under OpenAI's usage terms — they are usable for marketing campaigns, product photography, social media, and brand design. Always credit the original prompt author when adapting a prompt closely.

What are the best GPT Image 2 prompts for beginners?

Start with short, focused prompts that describe one subject, one clear style reference, and one lighting setup. The Profile / Avatar and Portrait & Photography categories on this page are the easiest entry points — most are 100–300 words and produce strong results without prompt engineering. Avoid stacking five styles in one prompt; pick one and iterate.

How do I write better GPT Image 2 prompts for portraits?

Specify camera type (35mm, medium format), lighting setup (side light, golden hour, harsh flash), film stock emulation (Fujifilm Pro 400H, Kodak Portra 400), and physical details like skin texture, fabric drape, and hair strands. Replace vague adjectives ("beautiful", "stunning") with concrete detail. The "How to write better GPT Image 2 prompts" section above has six concrete tips drawn from the top performers in this library.

How is this different from Nano Banana / Gemini image models?

GPT Image 2 is OpenAI; Nano Banana Pro is Google Gemini. They have different strengths: • GPT Image 2 — best for typography, dense layouts, posters, infographics, and structural instructions • Nano Banana Pro — best for consistent characters across images, 4K photoreal output, and reference-image editing The 2Slides workspace works with both. For posters and UI mockups, GPT Image 2 is the stronger choice; for character-consistent storyboards, Nano Banana Pro is.

How do I actually run a prompt?

Two ways. (1) Click "Use this prompt" on any card — the prompt loads into the generator panel at the top of this page; pick size + quality and hit Generate. We charge credits and call OpenAI for you. (2) Click "Copy" and use the prompt with your own OpenAI API key (model: gpt-image-2) or paste it into ChatGPT/any tool that wraps the same API.

Why does the price change with size and quality?

OpenAI bills gpt-image-2 by output pixels and quality tier. We mirror that pricing in credits: • 1024×1024 — Low 20 / Medium 100 / High 400 credits • 1024×1536 (vertical) — Low 30 / Medium 150 / High 600 credits • 1536×1024 (horizontal) — Low 30 / Medium 150 / High 600 credits Medium 1024×1024 is anchored at 100 credits — the same rate as our workspace Nano Banana 1K page generation, so pricing stays consistent across products.

What aspect ratios does GPT Image 2 support?

Three sizes via the Images API: 1024×1024 (1:1 square), 1024×1536 (2:3 portrait), and 1536×1024 (3:2 landscape). For social media posts and avatars use 1:1; for posters and IG stories use 2:3; for blog headers and YouTube thumbnails use 3:2.

What happens if generation fails?

Your credits are automatically refunded. The generation dialog will show the error reason (rate limit, content policy, or upstream issue) and offer a Retry button. We never charge for a failed generation — credits are deducted atomically only after OpenAI returns a valid image.

Are the YouMind JSON-template prompts ready to use?

Yes. The YouMind library uses a templating syntax `{argument name="x" default="y"}` for its Raycast extension. Our parser auto-resolves these placeholders to their default values before display, so the prompt you copy or run is directly compatible with the OpenAI Images API. No manual editing required.

How often is this list updated?

The library mirrors the upstream EvoLinkAI and YouMind repositories and refreshes at most once per hour. New community cases that the upstream curators add appear automatically, with no action needed on your side.