GPT Image 2thumbnailphotography3:2Reusable recipe

Creator Build Thumbnail

Generate a legible creator thumbnail with one tactile build object, a compact headline, and strong mobile contrast.

Open in Retake

Creator thumbnail with a tiny modular robot and Build Small headline
BUILD SMALL · modular robot · cobalt and red

Complete prompt

Copy the direction, then make it yours.

Create a landscape 3:2 creator-video thumbnail. Show {{BUILD_OBJECT}} as the clear hero, photographed close-up on a workshop table with energetic hands entering the frame. Use a bold {{PALETTE}} graphic background, one oversized exact headline "{{HEADLINE}}" occupying no more than one third of the image, and one directional arrow integrated as a clean shape. High contrast at mobile size, dramatic side lighting, tactile materials, believable shadows, no face close-up, no logos, no extra text, no watermark, no clickbait emoji.

Reusable variables

Swap the brief without losing the art direction.

{{BUILD_OBJECT}}

Build object

One visually clear object being built.

Default
a tiny modular robot
Examples
a tiny modular robot · a translucent neon radio prototype
{{PALETTE}}

Palette

A compact three-color palette.

Default
cobalt blue, cream, and red
Examples
cobalt blue, cream, and red · violet, black, and lime
{{HEADLINE}}

Headline

Two or three words that stay readable on mobile.

Default
BUILD SMALL
Examples
BUILD SMALL · MAKE IT REAL

More Retake previews

Same art direction, another filled brief.

Creator thumbnail with a translucent radio prototype and Make It Real headline
MAKE IT REAL · translucent radio · violet and lime
build object
a translucent neon radio prototype
headline
MAKE IT REAL
palette
violet, black, and lime

Original Retake direction

Clear prompt and preview provenance.

The subject, composition, copy, variables, and visual system are original Retake content. Every preview was generated as an original Retake Catalog asset during catalog preparation and reviewed against this exact prompt. These previews do not claim output from the target model. Source and license observations remain recorded in the internal Catalog for review.

Official model referenceGPT Image 2 modelOpenAI lists gpt-image-2 as its current image generation and editing model.